WO2024043009A1 - 信号処理方法および信号処理装置 - Google Patents
信号処理方法および信号処理装置 Download PDFInfo
- Publication number
- WO2024043009A1 WO2024043009A1 PCT/JP2023/027993 JP2023027993W WO2024043009A1 WO 2024043009 A1 WO2024043009 A1 WO 2024043009A1 JP 2023027993 W JP2023027993 W JP 2023027993W WO 2024043009 A1 WO2024043009 A1 WO 2024043009A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- image
- restoration
- parameter group
- restored
- compressed
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Ceased
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F3/00—Input arrangements for transferring data to be processed into a form capable of being handled by the computer; Output arrangements for transferring data from processing unit to output unit, e.g. interface arrangements
- G06F3/01—Input arrangements or combined input and output arrangements for interaction between user and computer
- G06F3/048—Interaction techniques based on graphical user interfaces [GUI]
- G06F3/0484—Interaction techniques based on graphical user interfaces [GUI] for the control of specific functions or operations, e.g. selecting or manipulating an object, an image or a displayed text element, setting a parameter value or selecting a range
- G06F3/04847—Interaction techniques to control parameter settings, e.g. interaction with sliders or dials
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T9/00—Image coding
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01J—MEASUREMENT OF INTENSITY, VELOCITY, SPECTRAL CONTENT, POLARISATION, PHASE OR PULSE CHARACTERISTICS OF INFRARED, VISIBLE OR ULTRAVIOLET LIGHT; COLORIMETRY; RADIATION PYROMETRY
- G01J3/00—Spectrometry; Spectrophotometry; Monochromators; Measuring colours
- G01J3/28—Investigating the spectrum
- G01J3/30—Measuring the intensity of spectral lines directly on the spectrum itself
- G01J3/36—Investigating two or more bands of a spectrum by separate detectors
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F3/00—Input arrangements for transferring data to be processed into a form capable of being handled by the computer; Output arrangements for transferring data from processing unit to output unit, e.g. interface arrangements
- G06F3/01—Input arrangements or combined input and output arrangements for interaction between user and computer
- G06F3/048—Interaction techniques based on graphical user interfaces [GUI]
- G06F3/0481—Interaction techniques based on graphical user interfaces [GUI] based on specific properties of the displayed interaction object or a metaphor-based environment, e.g. interaction with desktop elements like windows or icons, or assisted by a cursor's changing behaviour or appearance
- G06F3/0482—Interaction with lists of selectable items, e.g. menus
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F3/00—Input arrangements for transferring data to be processed into a form capable of being handled by the computer; Output arrangements for transferring data from processing unit to output unit, e.g. interface arrangements
- G06F3/01—Input arrangements or combined input and output arrangements for interaction between user and computer
- G06F3/048—Interaction techniques based on graphical user interfaces [GUI]
- G06F3/0484—Interaction techniques based on graphical user interfaces [GUI] for the control of specific functions or operations, e.g. selecting or manipulating an object, an image or a displayed text element, setting a parameter value or selecting a range
- G06F3/04842—Selection of displayed objects or displayed text elements
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F3/00—Input arrangements for transferring data to be processed into a form capable of being handled by the computer; Output arrangements for transferring data from processing unit to output unit, e.g. interface arrangements
- G06F3/01—Input arrangements or combined input and output arrangements for interaction between user and computer
- G06F3/048—Interaction techniques based on graphical user interfaces [GUI]
- G06F3/0484—Interaction techniques based on graphical user interfaces [GUI] for the control of specific functions or operations, e.g. selecting or manipulating an object, an image or a displayed text element, setting a parameter value or selecting a range
- G06F3/04845—Interaction techniques based on graphical user interfaces [GUI] for the control of specific functions or operations, e.g. selecting or manipulating an object, an image or a displayed text element, setting a parameter value or selecting a range for image manipulation, e.g. dragging, rotation, expansion or change of colour
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T1/00—General purpose image data processing
- G06T1/20—Processor architectures; Processor configuration, e.g. pipelining
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/25—Determination of region of interest [ROI] or a volume of interest [VOI]
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N23/00—Cameras or camera modules comprising electronic image sensors; Control thereof
- H04N23/60—Control of cameras or camera modules
Definitions
- the present disclosure relates to a signal processing method and a signal processing device.
- Compressive sensing is a technology that restores more data than the observed data by assuming that the data distribution of the observed object is sparse in a certain space (for example, frequency space).
- Compressed sensing can be applied to imaging systems that restore images containing more information from a small number of observed data.
- an optical filter may be used that has the function of encoding the light image, for example spatially or wavelength-wise.
- Such an imaging system can obtain a compressed image by imaging a subject through an optical filter, and can generate a restored image containing more information than the compressed image through calculation. This makes it possible to obtain various effects such as, for example, higher image resolution, multiple wavelengths, shorter imaging time, and higher sensitivity.
- Patent Document 1 discloses an example in which compressed sensing technology is applied to a hyperspectral camera that acquires images of multiple wavelength bands, each of which is a narrow band. According to the technology disclosed in Patent Document 1, a high-resolution, multi-wavelength hyperspectral image can be generated.
- Patent Document 2 discloses a super-resolution method that uses compressed sensing technology to generate high-resolution images from a small amount of observation information.
- Patent Document 3 discloses a method of restoring an image with a higher resolution than the acquired image by applying a convolutional neural network (CNN) to the acquired image.
- CNN convolutional neural network
- the present disclosure provides a technique for improving the efficiency or performance of processing for restoring a restored image with a larger amount of information from a compressed image.
- a method is a signal processing method performed using a computer.
- the method includes obtaining a compressed image including compressed information of a subject, and obtaining a first parameter group and a restoration matrix used in a first restoration process for generating a restoration target image from the compressed image. and generating the restoration target image based on the compressed image, the first parameter group, and the restoration matrix, and a second restoration process used in a second restoration process for generating a restored image from the compressed image. obtaining a parameter group; generating the restored image based on the compressed image, the second parameter group, and the restoration matrix; and generating the second restored image based on the restoration target image and the restored image. and correcting the parameter group.
- Generic or specific aspects of the present disclosure may be realized by a system, apparatus, method, integrated circuit, computer program, or recording medium such as a computer-readable recording disk, and the system, apparatus, method, integrated circuit, It may be realized by any combination of a computer program and a recording medium.
- the computer-readable recording medium may include a volatile recording medium or a non-volatile recording medium such as a CD-ROM (Compact Disc-Read Only Memory).
- a device may be composed of one or more devices. When the device is composed of two or more devices, the two or more devices may be placed within one device, or may be separately placed within two or more separate devices.
- “device” may refer not only to a device, but also to a system of devices.
- FIG. 1 is a flowchart illustrating a signal processing method according to an embodiment of the present disclosure.
- FIG. 2A is a diagram schematically showing a configuration example of an imaging system.
- FIG. 2B is a diagram illustrating a configuration example of an imaging device in which a filter array is placed apart from an image sensor.
- FIG. 2C is a diagram illustrating another configuration example of an imaging device in which a filter array is placed apart from an image sensor.
- FIG. 2D is a diagram illustrating still another configuration example of an imaging device in which a filter array is placed apart from an image sensor.
- FIG. 3A is a diagram schematically showing an example of a filter array.
- FIG. 1 is a flowchart illustrating a signal processing method according to an embodiment of the present disclosure.
- FIG. 2A is a diagram schematically showing a configuration example of an imaging system.
- FIG. 2B is a diagram illustrating a configuration example of an imaging device in which a filter array is placed apart from an image sensor.
- FIG. 3B is a diagram showing an example of the spatial distribution of light transmittance of each of a plurality of wavelength bands included in the target wavelength range.
- FIG. 3C is a diagram showing an example of the spectral transmittance of area A1 included in the filter array shown in FIG. 3A.
- FIG. 3D is a diagram showing an example of the spectral transmittance of area A2 included in the filter array shown in FIG. 3A.
- FIG. 4A is a diagram for explaining an example of the relationship between a target wavelength range W and a plurality of wavelength bands W 1 , W 2 , . . . , W N included therein.
- FIG. 4B is a diagram for explaining another example of the relationship between the target wavelength range W and the plurality of wavelength bands W 1 , W 2 , . . . , W N included therein.
- FIG. 5 is a block diagram showing a configuration example of a signal processing device.
- FIG. 6 is a conceptual diagram showing an example of the first restoration process and the second restoration process.
- FIG. 7 is a conceptual diagram showing another example of the first restoration process and the second restoration process.
- FIG. 8 is a flowchart illustrating an example of a method for adjusting the second parameter group.
- FIG. 9 is a diagram showing an example of a GUI that allows the user to confirm whether the restoration target image can be adopted.
- FIG. 9 is a diagram showing an example of a GUI that allows the user to confirm whether the restoration target image can be adopted.
- FIG. 10 is a diagram showing an example of a restoration target image and a restored image displayed on a display device.
- FIG. 11 is a flowchart showing a modification of the method shown in FIG.
- FIG. 12 is a diagram showing an example in which a partial area in an image is designated by the user.
- FIG. 13 is a diagram showing an example of a displayed warning.
- FIG. 14 is a diagram showing an example of a GUI for changing parameters.
- FIG. 15 is a diagram showing an example of a one-dimensional vector of n ⁇ m rows and one column based on image data that is two-dimensional data of n ⁇ m pixels.
- all or part of a circuit, unit, device, member, or section, or all or part of a functional block in a block diagram may be, for example, a semiconductor device, a semiconductor integrated circuit (IC), or an LSI (large scale integration). ) may be implemented by one or more electronic circuits.
- An LSI or IC may be integrated into one chip, or may be configured by combining a plurality of chips.
- functional blocks other than the memory element may be integrated into one chip.
- it is called LSI or IC, but the name changes depending on the degree of integration, and it may also be called system LSI, VLSI (very large scale integration), or ULSI (ultra large scale integration).
- Field Programmable Gate Arrays (FPGAs) which are programmed after the LSI is manufactured, or reconfigurable logic devices that can reconfigure the connections inside the LSI or set up circuit sections inside the LSI can also be used for the same purpose.
- the functions or operations of all or part of a circuit, unit, device, member, or section can be performed by software processing.
- the software is recorded on one or more non-transitory storage media, such as a ROM, an optical disk, or a hard disk drive, and when the software is executed by a processor, the functions specified in the software are performed. Executed by processing units and peripheral devices.
- the system or apparatus may include one or more non-transitory storage media on which software is recorded, a processing unit, and necessary hardware devices, such as interfaces.
- data or signals representing an image may be simply referred to as an "image.”
- Various algorithms can be applied to the process of restoring a restored image containing more information from a compressed image with less information. For example, various algorithms based on the aforementioned compressed sensing techniques or machine learning such as deep learning may be utilized. Each algorithm has unique characteristics. For example, when a certain algorithm is used, highly accurate restoration is possible, but the amount of calculation is large and the restoration process may take a long time. On the other hand, when other algorithms are used, restoration is possible in a short time, but the restoration accuracy may be poor.
- an imaging device equipped with an optical filter as disclosed in Patent Document 1
- Such imaging devices produce compressed images by sequentially imaging the product through an optical filter that encodes the image of light in terms of wavelength.
- a restored image (for example, a hyperspectral image) can be generated by applying computation using an algorithm such as compressed sensing or machine learning to the generated compressed image. Based on the generated restored image, it is possible to inspect whether there are any abnormalities in the product or whether there are any foreign substances mixed into the product.
- Such systems require real-time processing. Therefore, it is required to perform restoration processing in a short time using a high-speed algorithm.
- the present disclosure is based on the above consideration, and provides a technique for efficiently optimizing one or more parameters in a restoration processing algorithm actually used in a field such as an inspection.
- FIG. 1 is a flowchart illustrating a signal processing method according to an embodiment of the present disclosure.
- the signal processing method is performed by a computer.
- the signal processing method shown in FIG. 1 includes the following steps S11 to S16.
- S11 A compressed image containing compressed information of the subject is acquired.
- S12 A first parameter group and a restoration matrix used in the first restoration process for generating a restoration target image from the compressed image are acquired.
- S13 A restoration target image is generated based on the compressed image, the first parameter group, and the restoration matrix.
- S14 Obtain a second parameter group used in a second restoration process for generating a restored image from the compressed image.
- S15 A restored image is generated based on the compressed image, the second parameter group, and the restoration matrix.
- S16 Correct the second parameter group based on the restoration target image and the restoration image.
- the "compressed image” is an image with a relatively small amount of information obtained by imaging.
- the compressed image may be, for example, image data in which information of a plurality of wavelength bands is compressed as one image information, but is not limited thereto.
- the compressed image may be image data for restoring an MRI (Magnetic Resonance Imaging) image.
- the compressed image may be image data for generating a high resolution image.
- “Restoration target image” is data of an image that is a restoration target.
- the restoration target image is generated from the compressed image by a first restoration process using a first algorithm.
- the first algorithm for example, an algorithm that requires a large amount of calculation but has high restoration performance may be adopted.
- an algorithm that performs a restoration process based on compressed sensing may be adopted as the first algorithm.
- the first parameter group is set.
- the first parameter group includes one or more parameters.
- the first parameter group may include multiple parameters or a single parameter.
- the parameters included in the first parameter group are sometimes referred to as "first parameters.”
- the first parameter group may be set by the user or automatically by the system.
- the set values of the first parameter group may be stored in a storage medium such as a memory.
- the restoration target image may be simply referred to as a "target image.”
- a "restored image” is an image that is restored for purposes such as inspection or analysis.
- the restored image is generated from the compressed image by a second restoration process using a second algorithm.
- an algorithm with a smaller calculation load than the first algorithm may be adopted.
- an algorithm that is faster or uses less memory than the first algorithm may be adopted as the second algorithm.
- a second parameter group is set before the second restoration process is performed.
- the second parameter group includes one or more parameters.
- the second parameter group may include multiple parameters or a single parameter.
- the parameters included in the second parameter group are sometimes referred to as "second parameters.”
- the second parameter group may include more parameters than the first parameter group.
- the number of parameters in the second parameter group may be twice or more, five times or more, or ten times or more the number of parameters in the first parameter group.
- the second parameter group may include, for example, 10 or more, 30 or more, or 50 or more parameters.
- Restoration matrix is matrix data used in the first restoration process and the second restoration process.
- the restoration matrix may be stored in a storage medium such as a memory in the form of a table or the like. For this reason, the restoration matrix is sometimes referred to as a "restoration table.”
- the restoration matrix may be a matrix that reflects the characteristics of an optical filter used in imaging based on compressed sensing, for example.
- the correction of the second parameter group in step S16 may include correcting one or more parameters included in the second parameter group so that the restored image approaches the restoration target image.
- the correction of the second parameter group involves determining the error evaluation value between the target restored image and the restored image, and correcting one or more parameters included in the second parameter group so as to minimize the error evaluation value. may include.
- the second parameter group can be tuned so that the restored image approaches the restoration target image.
- the final second parameter group may be determined by repeating the process in step S15 and the process in step S16 multiple times. That is, the signal processing method corrects the second parameter group based on the restoration target image and the restored image, and generates the restored image using the corrected second parameter group, by repeating multiple times.
- the method may include determining a final second parameter group. Thereby, it is possible to optimize the second parameter group so that, for example, the restored image almost matches the restored target image.
- the second restoration process may include processing based on a learned model learned through machine learning, such as deep learning. Such processing is fast and can generate restored images in a short time. Algorithms based on machine learning require optimization of a large number of parameters. In this embodiment, parameters can be efficiently optimized based on a highly accurate restoration target image.
- the first restoration process does not need to include processing based on a learned model learned through machine learning.
- the first restoration process may include, for example, an iterative operation that minimizes or maximizes an evaluation function based on the compressed image and the restoration matrix.
- algorithms that perform such iterative calculations are capable of highly accurate restoration, they require a large amount of calculations and may not be able to generate a restored image in a short time. Therefore, the first algorithm that performs the first restoration process is not used in a real environment such as an inspection, but generates a restoration target image that is referred to to correct the second parameter group in the second algorithm used in the real environment. used for By correcting the second parameter group using the highly accurate restoration target image obtained by the first restoration processing, it is possible to improve the restoration performance by the second restoration processing.
- the compressed image may be an image in which spectral information of the subject is encoded.
- the compressed image may be an image in which information of a plurality of wavelength bands of a subject is compressed into one monochrome image.
- each of the restoration target image and the restoration image may include information on a plurality of images corresponding to a plurality of wavelength bands. This allows images of multiple wavelength bands (for example, hyperspectral images) to be restored from the compressed image.
- the method may further include displaying a graphical user interface (GUI) on the display device for the user to input the first group of parameters. This allows the user to set the first parameter group via the GUI.
- GUI graphical user interface
- the above method corrects the second parameter group based on the restoration target image and the restored image, and generates the restored image using the corrected second parameter group for a predetermined number of times unless an end condition is met. repeating the process; calculating an error evaluation value between the restored image and the restoration target image after a predetermined number of repetitions; and re-inputting the first parameter group when the error evaluation value is larger than a threshold;
- the method may further include displaying on the display device a GUI that prompts the user to change at least one of a predetermined number of times and a change in the termination condition. This allows the user to change the generation conditions for the restoration target image when the error evaluation value between the restored image and the restoration target image does not become less than or equal to the threshold value.
- Calculating the above error evaluation value involves extracting a first region in the restoration target image, extracting a second region corresponding to the first region in the restoration image, and comparing the first region and the second region. and determining the error evaluation value based on the difference.
- the first area and the second area may be determined based on, for example, an area designated by the user.
- the second parameter group can be corrected so as to reduce the error in the region extracted from the restored image and the restored target image.
- a signal processing device includes one or more processors and a memory storing a computer program executed by the one or more processors.
- the computer program causes the one or more processors to perform the signal processing method described above. That is, the computer program instructs the one or more processors to (a) obtain a compressed image including compressed information of a subject; and (b) generate a restoration target image from the compressed image. (c) generating the restoration target image based on the compressed image, the first parameter group, and the restoration matrix; (d) obtaining a second parameter group used in a second restoration process for generating a restored image from the compressed image; and (e) based on the compressed image, the second parameter group, and the restoration matrix. and (f) correcting the second parameter group based on the restoration target image and the restored image.
- Imaging system Next, a configuration example of an imaging system that may be used in an exemplary embodiment of the present disclosure will be described.
- FIG. 2A is a diagram schematically showing a configuration example of an imaging system.
- This imaging system includes an imaging device 100 and a signal processing device 200 (hereinafter simply referred to as "processing device 200").
- the imaging device 100 has the same configuration as the imaging device disclosed in Patent Document 1.
- the imaging device 100 includes an optical system 140, a filter array 110, and an image sensor 160.
- Optical system 140 and filter array 110 are arranged on the optical path of light incident from object 70, which is a subject.
- Filter array 110 in the example of FIG. 2A is placed between optical system 140 and image sensor 160.
- an apple is illustrated as an example of the target object 70.
- the object 70 is not limited to an apple, but may be any object.
- the image sensor 160 generates data of a compressed image 10 in which information of a plurality of wavelength bands is compressed as a two-dimensional monochrome image.
- the processing device 200 generates image data for each of a plurality of wavelength bands included in a predetermined target wavelength range based on the data of the compressed image 10 generated by the image sensor 160.
- This generated image data of a plurality of wavelength bands is sometimes referred to as a "hyperspectral (HS) data cube" or "hyperspectral image data.”
- HS hyperspectral
- the number of wavelength bands included in the target wavelength range is assumed to be N (N is an integer of 4 or more).
- restored images 20W 1 , 20W 2 the generated image data of a plurality of wavelength bands
- restored images 20W 1 , 20W 2 the generated image data of a plurality of wavelength bands
- data or signals representing an image that is, a collection of data or signals representing the pixel value of each pixel may be simply referred to as an "image.”
- the filter array 110 in this embodiment is an array of a plurality of light-transmitting filters arranged in rows and columns.
- the plurality of filters include a plurality of types of filters having different wavelength dependencies of spectral transmittance, that is, light transmittance.
- the filter array 110 modulates the intensity of incident light for each wavelength and outputs the modulated light. This process by filter array 110 is referred to as "encoding,” and filter array 110 is sometimes referred to as an "encoding element" or "encoding mask.”
- the filter array 110 is placed near or directly above the image sensor 160.
- “nearby” means that the image of light from the optical system 140 is close enough to be formed on the surface of the filter array 110 in a somewhat clear state.
- “Directly above” means that the two are so close that there is almost no gap between them.
- Filter array 110 and image sensor 160 may be integrated.
- the optical system 140 includes at least one lens. Although optical system 140 is shown as one lens in FIG. 2A, optical system 140 may be a combination of multiple lenses. Optical system 140 forms an image on the imaging surface of image sensor 160 via filter array 110.
- FIGS. 2B to 2D are diagrams illustrating a configuration example of the imaging device 100 in which the filter array 110 is placed apart from the image sensor 160.
- filter array 110 is arranged between optical system 140 and image sensor 160 and at a location away from image sensor 160.
- filter array 110 is placed between object 70 and optical system 140.
- the imaging device 100 includes two optical systems 140A and 140B, with the filter array 110 disposed between them.
- an optical system including one or more lenses may be placed between the filter array 110 and the image sensor 160.
- the image sensor 160 is a monochrome photodetection device that has a plurality of photodetection elements (also referred to as "pixels" in this specification) arranged in a two-dimensional plane.
- the image sensor 160 may be, for example, a CCD (Charge-Coupled Device), a CMOS (Complementary Metal Oxide Semiconductor) sensor, or an infrared array sensor.
- the photodetecting element includes, for example, a photodiode.
- Image sensor 160 does not necessarily have to be a monochrome type sensor. For example, use a color type sensor with red (R)/green (G)/blue (B), R/G/B/infrared (IR), or R/G/B/transparent (W) filters.
- the wavelength range to be acquired may be arbitrarily determined, and is not limited to the visible wavelength range, but may be an ultraviolet, near-infrared, mid-infrared, or far-infrared wavelength range.
- Processing device 200 may be a computer with one or more processors and one or more storage media, such as memory.
- the processing device 200 generates data of a plurality of restored images 20W 1 , 20W 2 , . . . 20W N based on the compressed image 10 acquired by the image sensor 160.
- the processing device 200 may be incorporated into the imaging device 100.
- FIG. 3A is a diagram schematically showing an example of the filter array 110.
- Filter array 110 has a plurality of regions arranged in a two-dimensional plane. In this specification, the area may be referred to as a "cell.”
- Optical filters having individually set spectral transmittances are arranged in each region.
- the spectral transmittance is expressed by a function T( ⁇ ), where the wavelength of incident light is ⁇ .
- the spectral transmittance T( ⁇ ) can take a value of 0 or more and 1 or less.
- the filter array 110 has 48 rectangular areas arranged in 6 rows and 8 columns. This is just an example, and in actual applications, more areas may be provided. The number may be approximately the same as the number of pixels of the image sensor 160, for example. The number of filters included in the filter array 110 is determined depending on the application, for example, in the range of several tens to tens of millions.
- FIG. 3B is a diagram showing an example of the spatial distribution of the transmittance of light in each of a plurality of wavelength bands W 1 , W 2 , . . . , W N included in the target wavelength range.
- the difference in shading in each region represents the difference in transmittance. The lighter the area, the higher the transmittance, and the darker the area, the lower the transmittance.
- the spatial distribution of light transmittance differs depending on the wavelength band.
- FIGS. 3C and 3D are diagrams showing examples of spectral transmittances of area A1 and area A2 included in filter array 110 shown in FIG. 3A, respectively.
- the spectral transmittance of area A1 and the spectral transmittance of area A2 are different from each other. In this way, the spectral transmittance of the filter array 110 differs depending on the region. However, the spectral transmittances of all regions do not necessarily have to be different. In the filter array 110, at least some of the plurality of regions have different spectral transmittances.
- Filter array 110 includes two or more filters with different spectral transmittances.
- the number of spectral transmittance patterns in the plurality of regions included in the filter array 110 may be equal to or greater than the number N of wavelength bands included in the target wavelength range.
- the filter array 110 may be designed such that half or more of the regions have different spectral transmittances.
- the target wavelength range W can be set to various ranges depending on the application.
- the target wavelength range W may be, for example, a visible light wavelength range from about 400 nm to about 700 nm, a near-infrared wavelength range from about 700 nm to about 2500 nm, or a near-ultraviolet wavelength range from about 10 nm to about 400 nm.
- the target wavelength range W may be a wavelength range such as mid-infrared or far-infrared. In this way, the wavelength range used is not limited to the visible light range.
- "light” refers to all radiation including not only visible light but also infrared and ultraviolet light.
- N is an arbitrary integer greater than or equal to 4, and the wavelength ranges obtained by dividing the target wavelength range W into N equal parts are designated as wavelength band W 1 , wavelength band W 2 , ..., wavelength band W N There is.
- the plurality of wavelength bands included in the target wavelength range W may be set arbitrarily.
- the bandwidth may be made non-uniform depending on the wavelength band.
- the bandwidth differs depending on the wavelength band, and there is a gap between two adjacent wavelength bands. In this way, the plurality of wavelength bands can be determined arbitrarily.
- a gray scale transmittance distribution is assumed in which the transmittance of each region can take any value between 0 and 1.
- a binary scale transmittance distribution may be adopted in which the transmittance of each region can take either a value of approximately 0 or approximately 1.
- each region transmits most of the light in at least two of the wavelength ranges included in the target wavelength range, and transmits most of the light in the remaining wavelength ranges. I won't let you.
- "most" refers to approximately 80% or more.
- a part of all the cells may be replaced with a transparent area.
- a transparent region transmits light in all wavelength bands W 1 to W N included in the target wavelength range W with a similar high transmittance, for example, 80% or more.
- the plurality of transparent regions may be arranged in a checkerboard pattern, for example. That is, in the two arrangement directions of the plurality of regions in the filter array 110, regions whose light transmittances differ depending on the wavelength and transparent regions may be arranged alternately.
- Such data indicating the spatial distribution of spectral transmittance of the filter array 110 may be obtained in advance based on design data or actual measurement calibration, and stored in a storage medium included in the processing device 200. This data is used for calculation processing described later.
- the filter array 110 may be configured using, for example, a multilayer film, an organic material, a diffraction grating structure, or a fine structure containing metal.
- a multilayer film for example, a dielectric multilayer film or a multilayer film including a metal layer can be used.
- at least one of the thickness, material, and lamination order of each multilayer film is formed to be different for each cell. This makes it possible to achieve different spectral characteristics depending on the cell.
- By using a multilayer film sharp rises and falls in spectral transmittance can be realized.
- a structure using organic materials can be realized by containing different pigments or dyes depending on the cell, or by stacking different types of materials.
- a configuration using a diffraction grating structure can be realized by providing diffraction structures with different diffraction pitches or depths for each cell.
- a fine structure containing metal it can be manufactured using spectroscopy based on the plasmon effect.
- the processing device 200 reconstructs a multi-wavelength hyperspectral image 20 based on the compressed image 10 output from the image sensor 160 and the spatial distribution characteristics of transmittance for each wavelength of the filter array 110.
- multiple wavelengths means, for example, more wavelength ranges than the three color wavelength ranges of RGB (red, green, and blue) acquired by a normal color camera.
- the number of wavelength ranges can be, for example, about 4 to 100.
- the number of wavelength ranges is referred to as the "number of bands.” Depending on the application, the number of bands may exceed 100.
- the data to be obtained is the data of the hyperspectral image 20, and this data is designated as f.
- f is the image data f 1 of the image corresponding to the wavelength band W 1
- the image data f 2 of the image corresponding to the wavelength band W 2 ...
- the image corresponding to the wavelength band W N This is data obtained by integrating the image data fN of .
- the horizontal direction of the image is the x direction
- the vertical direction of the image is the y direction.
- each of image data f 1 , image data f 2 , ..., image data f N has n ⁇ m pixels. This is two-dimensional data. Therefore, the data f is three-dimensional data with the number of elements n ⁇ m ⁇ N. This three-dimensional data is referred to as "hyperspectral image data" or "hyperspectral data cube.”
- the image data g of the compressed image 10 obtained by being encoded and multiplexed by the filter array 110 is two-dimensional data of n ⁇ m pixels.
- Image data g can be expressed by the following equation (1).
- g included in formula (1) is a one-dimensional vector of n ⁇ m rows and one column based on the image data g, which is the two-dimensional data of n ⁇ m pixels described above.
- f 1 included in equation (1) is a one-dimensional vector of n x m rows and 1 column based on the image data f 1 which is two-dimensional data of n x m pixels described above, and is included in equation (1).
- f2 is a one-dimensional vector of n ⁇ m rows and one column based on the image data f2 which is two-dimensional data of n ⁇ m pixels described above
- f N included in equation (1) is , is a one-dimensional vector of n ⁇ m rows and one column based on the image data fN , which is the two-dimensional data of n ⁇ m pixels described above.
- FIG. 15 is a diagram showing an example of a one-dimensional vector of n ⁇ m rows and one column based on image data that is two-dimensional data of n ⁇ m pixels.
- f included in equation (1) is a one-dimensional vector of n ⁇ m ⁇ N rows and 1 column.
- the matrix H is obtained by encoding and intensity modulating each component f 1 , f 2 , ..., f N of f with different encoding information (also referred to as "mask information") for each wavelength band, and adding them. Represents a transformation. Therefore, H is a matrix with n ⁇ m rows and n ⁇ m ⁇ N columns. This matrix H is sometimes called a "restoration matrix.”
- the processing device 200 uses the redundancy of the image included in the data f to find a solution using a compressed sensing technique. Specifically, the desired data f is estimated by solving Equation (2) below.
- f' represents the estimated data of f.
- the first term in parentheses in the above equation represents the amount of deviation between the estimation result Hf and the acquired data g, a so-called residual term.
- the second term in parentheses is a regularization term or stabilization term.
- Equation (2) means finding f that minimizes the sum of the first term and the second term.
- the function in parentheses in equation (2) is called an evaluation function.
- the processing device 200 can converge the solution through recursive iterative operations and calculate f that minimizes the evaluation function as the final solution f'.
- the first term in parentheses in Equation (2) means an operation to calculate the sum of squares of the difference between the acquired data g and Hf obtained by converting f in the estimation process using matrix H.
- the second term ⁇ (f) is a constraint in the regularization of f, and is a function reflecting the sparse information of the estimated data. This function has the effect of smoothing or stabilizing the estimated data.
- the regularization term may be represented by, for example, a discrete cosine transform (DCT), a wavelet transform, a Fourier transform, or a total variation (TV) of f. For example, when total variation is used, stable estimated data can be obtained that suppresses the influence of noise on the observed data g.
- DCT discrete cosine transform
- TV total variation
- the sparsity of the object 70 in the space of each regularization term differs depending on the texture of the object 70.
- a regularization term may be selected that makes the texture of the object 70 sparser in the space of regularization terms.
- multiple regularization terms may be included in the calculation.
- ⁇ is a weighting coefficient. The larger the weighting coefficient ⁇ , the greater the amount of redundant data to be reduced, and the higher the compression ratio. The smaller the weighting coefficient ⁇ , the weaker the convergence to a solution.
- the weighting coefficient ⁇ is set to an appropriate value that allows f to converge to some extent and not cause overcompression.
- the image encoded by the filter array 110 is acquired in a blurred state on the imaging surface of the image sensor 160. Therefore, the hyperspectral image 20 can be reconstructed by retaining this blur information in advance and reflecting the blur information in the matrix H described above.
- the blur information is expressed by a point spread function (PSF).
- PSF is a function that defines the degree of spread of a point image to surrounding pixels. For example, when a point image corresponding to one pixel on an image spreads to an area of k ⁇ k pixels around that pixel due to blur, the PSF is a group of coefficients that indicates the influence on the pixel value of each pixel in that area. , that is, can be defined as a matrix.
- the hyperspectral image 20 can be reconstructed by reflecting the effect of blurring of the coding pattern due to PSF on the matrix H.
- a position may be selected where the coding pattern of filter array 110 is too diffused and disappears.
- the hyperspectral image 20 can be restored from the compressed image 10 acquired by the image sensor 160.
- the process of generating a restored image from a compressed image is not limited to the algorithm using compressed sensing based on the above-mentioned equation (2), but can be executed using an algorithm using machine learning.
- Algorithms using machine learning include, for example, algorithms that apply a trained model learned by deep learning to a compressed image to generate a restored image. By using such an algorithm, the time required for the restoration process can be shortened.
- parameter optimization is required. Therefore, a method may be used in which each parameter in a machine learning algorithm is optimized based on a restoration target image generated using an algorithm based on compressed sensing. Thereby, parameter optimization can be performed efficiently.
- FIG. 5 is a block diagram showing a configuration example of the processing device 200.
- the processing device 200 shown in FIG. 5 is used in combination with the imaging device 100 and the terminal device 500.
- Imaging device 100 may be the hyperspectral imaging device described above.
- the terminal device 500 is a computer for executing various operations related to parameter optimization processing for generating a restored image.
- Terminal device 500 includes an input device 510 and a display device 520.
- Processing device 200 includes one or more processors 210, such as a CPU or GPU, and one or more memories 250.
- Memory 250 stores computer programs executed by processor 210 and various data generated by processor 210.
- the computer program causes processor 210 to perform the signal processing method illustrated in FIG. That is, the processor 210 executes the following process by executing a computer program.
- - Acquire the compressed image generated by the imaging device 100.
- - Acquire the first parameter group and the restoration matrix (corresponding to the matrix H described above) from the memory 250.
- - Generate a restoration target image based on the compressed image, the first parameter group, and the restoration matrix.
- - Obtaining a second parameter group from memory 250; - Generate a restored image based on the compressed image, the second parameter group, and the restoration matrix.
- Correct the second parameter group based on the restoration target image and the restoration image.
- the processor 210 has the functions of a restoration target image generation section 212, an image restoration section 214, and a parameter optimization section 216. These functions can be realized by software processing. Note that in this embodiment, one processor 210 executes all processes of the restoration target image generation section 212, the image restoration section 214, and the parameter optimization section 216, but this is only an example. These processes may be performed by multiple hardware (eg, circuits or computers). For example, the processing of restoration target image generation, image restoration, and parameter optimization may be performed by a plurality of computers connected to each other via a network.
- the restoration target image generation unit 212 executes a first restoration process based on the compressed image generated by the imaging device 100, the first parameter group and restoration matrix stored in the memory 250, and generates a restoration target image.
- the first algorithm is used for the first restoration process.
- the generated restoration target image is output to the display device 520 and the restoration parameter optimization unit 216.
- the image restoration unit 214 generates a restored image by performing a second restoration process based on the compressed image, the second parameter group, and the restoration matrix. A second algorithm is used for the second restoration process.
- the generated restored image is output to the display device 520 and the parameter optimization unit 216.
- the parameter optimization unit 216 corrects the second parameter group based on the restoration target image and the restoration image, and outputs the corrected second parameter group to the image restoration unit 214.
- the restoration parameter optimization unit 216 corrects the second parameter group so that the difference between the restored image and the restoration target image becomes smaller.
- the process of correcting the second parameter group based on the restoration target image and the restored image and the process of generating the restored image using the corrected second parameter group are repeated a predetermined number of times, and the final second parameter group is The value is determined. This optimizes the second parameter group.
- each of the restoration target image and the restoration image includes a plurality of images corresponding to a plurality of wavelength bands.
- the plurality of wavelength bands may include four or more bands with relatively narrow bandwidths, such as a band with a wavelength of 400-410 nm, a band with a wavelength of 410-420 nm, for example.
- Correcting the second parameter group so that the restored image approaches the restoration target image means correcting the second parameter group so that the image of each band in the restored image approaches the image of the corresponding band in the restoration target image. It may include correcting one or more values corresponding to one or more parameters.
- an image with a wavelength of 400-410 nm in the restored image approaches an image with a wavelength of 400-410 nm in the restored target image
- an image with a wavelength of 410-420 nm in the restored image approaches an image with a wavelength of 410-420 nm in the restored target image.
- a second set of parameters may be corrected.
- the magnitude of the difference between the restored image and the restored target image can be evaluated based on the error evaluation value.
- the error evaluation value can be calculated by, for example, calculating the error for each wavelength band image that has a one-to-one correspondence between the restored image and the restored target image, and then summing or averaging these errors. It can be calculated. That is, minimizing the error evaluation value may include minimizing the sum or average value of errors for each of the plurality of wavelength bands.
- the display device 520 displays the restoration target image and the restoration image that are generated in the process of optimizing the second parameter group.
- the input device 510 may include a device such as a keyboard or mouse used by the user to set various setting items such as the first parameter group.
- FIG. 6 is a conceptual diagram showing an example of the first restoration process and the second restoration process.
- the first restoration process in this example uses the first algorithm based on compressed sensing, and generates the restoration target image by the recursive iterative operation shown in equation (2) above.
- the first parameter group in the first algorithm may include several parameters, such as the number of iterations and the regularization coefficient ⁇ .
- the initial values of the first parameter group are stored in the memory 250 in advance.
- the first parameter group can also be set by the user using the input device 510.
- the first restoration process includes an iterative operation that minimizes the evaluation function based on the compressed image and the restoration matrix, that is, the function in parentheses on the right side of equation (2).
- the first restoration process will include an iterative operation that maximizes the evaluation function based on the compressed image and the restoration matrix.
- the second restoration process in the example of FIG. 6 generates a restored image using a second algorithm based on machine learning such as deep learning. Then, the second parameter group is corrected so as to minimize the error evaluation value between the restored image and the restored target image.
- the error evaluation value may be, for example, the sum or average value of the square or absolute value of the difference in each pixel value in the region of the subject of interest between the restored image and the restored target image.
- deep learning a plurality of nodes are provided in each of an input layer, a plurality of hidden layers, and an output layer, and each node has a unique weight.
- the second parameter group may include weights of each of the plurality of nodes in the deep learning algorithm and hyperparameters.
- Hyperparameters may include, for example, the number of learning times, the learning rate, and parameters specifying the learning algorithm.
- the second parameter group may include, for example, hundreds or more parameters. When the second parameter group is corrected, the weight of each of the plurality of nodes, the number of times of learning, the learning rate, and/or the algorithm that specifies the learning algorithm may be corrected.
- learning in the second restoration process that is, optimization of the second parameter group
- learning in the second restoration process is performed using the restoration target image generated in the first restoration process as learning data.
- the second parameter group can be efficiently optimized.
- the second restoration process does not necessarily need to be performed by a machine learning algorithm.
- the second restoration process may be performed by an algorithm based on compressed sensing.
- FIG. 7 is a conceptual diagram showing another example of the first restoration process and the second restoration process.
- both the first restoration process and the second restoration process are performed using an algorithm based on compressed sensing.
- the first restoration process may be executed, for example, by a first algorithm that has high restoration performance but has a high computational load.
- the second restoration process may be performed by a second algorithm that has a lower computational load than the first algorithm but requires setting more parameters.
- an iterative operation is performed based on an iterative operation function ⁇ 1 that depends on n first parameters p 1 , . . . , p n .
- an iterative operation is performed based on an iterative operation function ⁇ 2 that depends on m second parameters q 1 , . . . , q m that are larger than n.
- the restored target image generated in the k-th iterative calculation is expressed as f 1 k
- the restored image generated in the k-th iterative calculation is expressed as f 2 k .
- p n can be set by the user, for example. Since m is larger than n, adjusting the second parameters q 1 , . . . , q m is more difficult than adjusting the first parameters p 1 , . In this embodiment, the second parameters q 1 , ..., q m are optimized so that the restored image generated in the second restoration process approaches the restoration target image generated in the first restoration process. . Thereby, the second parameter group can be efficiently optimized.
- the second algorithm is selected from a different algorithm than the first algorithm.
- the second algorithm may be selected from algorithms that have some advantage over the first algorithm, such as shorter computation time, higher memory efficiency, or higher restoration accuracy.
- FIG. 8 is a flowchart illustrating an example of a method for adjusting the second parameter group. The method shown in FIG. 8 is executed by processor 210 of processing device 200.
- step S101 the processor 210 acquires the compressed image generated by the imaging device 100.
- the processor 210 may obtain the compressed image directly from the imaging device 100 or may obtain the compressed image via a storage medium such as the memory 250.
- step S102 the processor 210 obtains a restoration matrix from the memory 250.
- the restoration matrix is generated in advance and stored in memory 250.
- Step S102 may be performed before step S101, or may be performed simultaneously with step S101.
- step S103 the processor 210 obtains the values of the first parameter group from the memory 250.
- the processor 210 obtains the values of the set first parameter group.
- step S104 the processor 210 generates a restoration target image based on the compressed image, the restoration matrix, and the first parameter group. This process corresponds to the first restoration process described above, and is executed using the first algorithm. If the first algorithm is an algorithm that performs iterative calculations based on compressed sensing, a restored target image is generated by performing iterative calculations a preset number of times.
- step S105 it is determined whether the restoration target image is a desired image. This determination may be made based on the user's operation using the input device 510. For example, the processor 210 causes the display device 520 to display the generated restoration target image and a GUI that allows the user to confirm whether or not to adopt the restoration target image, and when the user approves the adoption of the restoration target image, , it may be determined that the restoration target image is a desired image.
- FIG. 9 is a diagram showing an example of a GUI that allows the user to confirm whether the restoration target image can be adopted.
- the generated restoration target image 521, the number of iterations and regularization coefficient that are the first parameters, and the GUI 522 are displayed on the display device 520.
- the GUI 522 includes "OK" and "parameter reset” buttons.
- the user presses the "OK” button the displayed restoration target image is adopted and the process proceeds to step S106.
- the user presses the "parameter reset” button the user can reset the first parameter. In that case, the process returns to step S103, and the processor 210 generates the restoration target image again using the reset first parameter. Thereafter, the processes from steps S103 to S105 are repeated until it is determined in step S105 that the desired image has been obtained.
- step S106 the processor 210 sets the second parameter group to the initial values stored in the memory 250.
- step S107 the processor 210 generates a restored image based on the compressed image, the restoration matrix, and the second parameter group.
- This process corresponds to the second restoration process described above, and is executed using the second algorithm. If the second algorithm is an algorithm that performs iterative calculations based on compressed sensing, a restored image is generated by performing iterative calculations a preset number of times.
- step S108 the processor 210 compares the restored image with the restored target image and evaluates the error. For example, the processor 210 determines the error evaluation value using an error evaluation function that indicates the magnitude of the difference between the restored image and the restoration target image.
- the error evaluation function for example, mean squared error (MSE) can be used. MSE is expressed by the following equation (3).
- n and m are the number of vertical and horizontal pixels of the image, respectively, f i,j is the pixel value of the i row, j column of the correct image, and I i,j is the i row, j column of the estimated restored image. represents the pixel value of
- the error is not limited to MSE, but also other errors such as Root MSE (RMSE), Peak-Signal-to-Noise-Ratio (PSNR), Mean Absolute Error (MAE), Structural Similarity (SSMI), or spectral angle. It can also be expressed as an evaluation index.
- each of the restoration target image and the restoration image includes a plurality of images corresponding to a plurality of wavelength bands.
- a plurality of images such as an image corresponding to a wavelength band of 400-410 nm and an image corresponding to a wavelength band of 410-420 nm, can be restored in each of the first restoration process and the second restoration process.
- an error evaluation function such as MSE may be calculated for each corresponding band between the restored image and the restored target image.
- the error evaluation value can be determined by summing or averaging the values of the error evaluation functions calculated for each band.
- the plurality of restoration target images include a first restoration target image corresponding to wavelength band W1 , a second restoration target image corresponding to wavelength band W2 , ..., an Nth restoration target image corresponding to wavelength band WN . May include.
- the plurality of restored images may include a first restored image corresponding to wavelength band W1 , a second restored image corresponding to wavelength band W2 , . . . an Nth restored image corresponding to wavelength band WN .
- the error evaluation values between the plurality of restoration target images and the plurality of restoration images include a first error evaluation value between the first restoration target image and the first restoration image, a second error evaluation value between the second restoration target image and the second restoration image, ..., may be determined based on the Nth restoration target image and the Nth error evaluation value of the Nth restoration image.
- the first error evaluation value is the MSE of the first restoration target image and the first restoration image
- the second error evaluation value is the MSE of the second restoration target image and the second restoration image
- the Nth error evaluation value is the Nth error evaluation value
- the MSE of the restoration target image and the Nth restoration image may be used.
- step S109 the processor 210 updates the second parameter group so that the error between the restored image and the restored target image becomes smaller.
- the second parameter group can be updated using a method such as gradient descent or Bayesian optimization to reduce the error evaluation value.
- step S110 the processor 210 determines whether a preset loop termination condition is satisfied.
- the loop termination condition may be, for example, that the optimization loop from steps S107 to S109 has been repeated a predetermined number of times, or that the error evaluation value between the restored image and the restored target image has become smaller than a threshold value. If the loop end condition is not satisfied, the process returns to step S107. If the loop end condition is satisfied, the process ends.
- the processor 210 may display the restored image in the generation process on the display device 520 together with the restored target image. This allows the user to compare the restored image and the restored target image and check whether the second parameter group has been optimized without any problem.
- FIG. 10 is a diagram showing an example of a restoration target image 521 and a restoration image 523 displayed on the display device 520.
- a restoration target image 521 and a restoration image 523 whose second parameter group has been optimized are displayed side by side.
- the first parameter for generating the restoration target image 521 and the termination condition for the optimization loop of the process for generating the restoration image 523 are also displayed.
- the first parameter and termination condition can be set by the user using the input device 510.
- the first parameter includes the number of iterative operations and a regularization coefficient
- the optimization loop termination condition includes the maximum number of loops and an error tolerance.
- the user can view the displayed restoration target image 521 and restoration image 523, change the parameters or the termination conditions of the optimization loop as needed, and have the processor 210 execute the restoration process again.
- FIG. 11 is a flowchart showing a modification of the method shown in FIG. 8.
- the flowchart shown in FIG. 11 has the points that step S206 is added between step S105 and step S106, step S108 is replaced with step S208, and steps S211 and S212 are executed when the determination is Yes in step S110. This is different from the flowchart shown in FIG. Hereinafter, points different from the example of FIG. 8 will be explained.
- step S206 the processor 210 extracts one or more regions from the restoration target image.
- Processor 210 may be configured, for example, to extract one or more regions specified by a user.
- the user specifies, for example, an area showing an important object that needs to be inspected or analyzed.
- FIG. 12 is a diagram showing an example in which a part of the area is specified by the user. In this example, an area A in which a specific subject is shown is specified. Area B that is not specified is processed as a background.
- the processor 210 may extract a region in which an important subject is estimated to exist from the restoration target image by image processing without depending on a user's operation.
- steps S106 and S107 are executed to generate a restored image.
- the processor 210 compares the restored image with the restored target image and evaluates the error. At this time, the processor 210 weights the evaluation value for each area so that the restoration accuracy of the designated area is improved. For example, in the example of FIG. 12, the weight of the evaluation value of the designated area is made relatively large, and the weight of the evaluation value of the unspecified area is made relatively small.
- the error evaluation value can be expressed by the following formula, where a and b are coefficients (a>b).
- Evaluation value a ⁇ evaluation value of area A+b ⁇ evaluation value of area B.
- the error in area B does not affect the evaluation value much, and the error in area A greatly affects the evaluation value.
- Coefficient a may be set to a larger value than coefficient b, for example 1.5 times, twice, or more than coefficient b.
- step S110 after it is determined in step S110 that the loop end condition is satisfied, it is determined whether the error evaluation value between the restored image and the restored target image is larger than a threshold (step S211). If this determination is Yes, the processor 210 causes the display device 520 to display a warning prompting the user to regenerate the target image (for example, re-input the first parameter group) or change the optimization loop termination conditions (step S212). .
- FIG. 13 is a diagram showing an example of a displayed warning.
- the message "The error in the restored image exceeds the threshold. Please change the parameters.” is displayed.
- the OK button the screen changes to a screen for changing parameters.
- FIG. 14 is a diagram showing an example of a GUI for changing parameters.
- an error image is displayed in addition to the target image and the optimized restored image.
- the error image is an image that indicates the degree of matching between the target image and the restored image, and may be, for example, a difference image, a squared error image, or a spectral angle distribution image.
- each of the target image and the restored image includes a plurality of images corresponding to a plurality of wavelength bands, but only an image corresponding to one wavelength band is illustrated in FIG. has been done. Images corresponding to each of a plurality of wavelength bands may be displayed.
- the user can change the first parameter for generating the target image and the conditions for terminating the optimization loop of the restored image while viewing the error image. Thereby, the first parameter and the optimization loop termination condition can be changed so that the error in the important subject area is as small as possible.
- the processor 210 corrects the second parameter group based on the restoration target image and the restored image, and generates the restored image using the corrected second parameter group. , is repeated a predetermined number of times unless the termination condition is met.
- the processor 210 calculates an error evaluation value between the restored image and the restored target image after a predetermined number of repetitions have been performed. When the error evaluation value is larger than the threshold, the processor 210 causes the display device 520 to display a GUI that prompts the user to at least one of re-entering the first parameter group, changing the predetermined number of times, and changing the termination condition. .
- the restoration target image can be regenerated with more appropriate values of the first parameter group, the above-mentioned predetermined number of times can be changed, or the termination condition can be changed. This makes it possible to generate a more suitable restored image.
- the processor 210 also extracts a first region in the restoration target image, extracts a second region corresponding to the first region in the restoration image, and calculates an error evaluation value based on the difference between the first region and the second region. decide.
- the first area may be specified by the user, for example. This makes it possible to optimize the second parameter group so that, for example, a restored image with small errors in areas of high importance is generated.
- a hyperspectral image is restored from a compressed image
- the scope of application of the technology of the present disclosure is not limited to the use of restoring a hyperspectral image.
- the technology of the present disclosure is applied to applications such as restoring a higher-resolution restored image from a low-resolution compressed image, restoring an MRI image from a compressed image, or restoring a three-dimensional image from a compressed image. be able to.
- a signal processing method performed using a computer comprising: obtaining a compressed image containing compressed information of a subject; obtaining a first parameter group and a restoration matrix used in a first restoration process for generating a restoration target image from the compressed image; generating the restoration target image based on the compressed image, the first parameter group, and the restoration matrix; obtaining a second parameter group used in a second restoration process for generating a restored image from the compressed image; generating the restored image based on the compressed image, the second parameter group, and the restoration matrix; correcting the second parameter group based on the restoration target image and the restoration image; method including.
- the second parameter group can be corrected appropriately and efficiently. As a result, the quality of the restored image generated by the second restoration process can be improved.
- Correcting the second parameter group means correcting one or more values corresponding to one or more parameters included in the second parameter group so that the restored image approaches the restoration target image.
- the second parameter group is appropriately corrected, making it possible to generate a restored image close to the restoration target image.
- Correcting the second parameter group includes: determining an error evaluation value between the target restored image and the restored image; correcting one or more values corresponding to one or more parameters included in the second parameter group so as to minimize the error evaluation value;
- the method according to technique 3 comprising:
- the second parameter group can be corrected more appropriately, and a restored image with small errors can be generated.
- the first restoration process is performed by a first algorithm
- the second restoration process is performed by a second algorithm
- the second algorithm is an algorithm with a lower computational load than the first algorithm
- a restored image can be generated in a shorter time by using the second algorithm, which has a smaller calculation load than the first algorithm.
- the first restoration process does not include a process based on a learned model learned through machine learning
- the second restoration process includes a process based on a learned model learned through machine learning. The method described in Technique 5.
- a restored image can be generated in a shorter time by the second restoration process that includes processing based on the learned model.
- the compressed image is an image in which spectral information of the subject is encoded,
- Each of the restoration target image and the restoration image includes information on a plurality of images corresponding to a plurality of wavelength bands.
- the method according to any one of techniques 1 to 7.
- the second parameter group can be corrected to more appropriate values, so the accuracy of the restored image can be improved.
- the user can adjust the first parameter group, so a more appropriate restoration target image can be generated.
- the second parameter group it becomes possible to correct the second parameter group to more appropriate values, and the accuracy of the restored image can be improved.
- the second parameter group can be corrected to more appropriate values. As a result, it becomes possible to generate a more accurate restored image.
- a signal processing device one or more processors; a memory storing a computer program executed by the one or more processors; Equipped with The computer program causes the one or more processors to: obtaining a compressed image containing compressed information of a subject; obtaining a first parameter group and a restoration matrix used in a first restoration process for generating a restoration target image from the compressed image; generating the restoration target image based on the compressed image, the first parameter group, and the restoration matrix; obtaining a second parameter group used in a second restoration process for generating a restored image from the compressed image; generating the restored image based on the compressed image, the second parameter group, and the restoration matrix; correcting the second parameter group based on the restoration target image and the restoration image; A signal processing device that executes.
- the second parameter group can be corrected appropriately and efficiently. As a result, the quality of the restored image generated by the second restoration process can be improved.
- the method according to the first item is a method performed by a computer, comprising: The method includes: a computer (a) receiving a first image of a subject imaged through a filter array including four or more filters, the four or more filters having a one-to-one correspondence with four or more spectral spectra; the four or more spectra are different from each other, (b) Based on the restoration matrix corresponding to the filter array and the plurality of pixel values of the first image, a plurality of first pixel values of the first restoration target image corresponding to the first wavelength range, ⁇ , nth wavelength; determining a plurality of n-th pixel values of the n-th restoration target image corresponding to the area; (c) determining a plurality of weight values of a neural network based on the plurality of pixel values of the first image, the plurality of first pixel values, ⁇ , the plurality of nth pixel values; each of the plurality of weight values corresponds to two directly connected units, and the neural network includes the
- the neural network uses the plurality of input values to generate an image corresponding to n wavelength range,
- the plurality of input values are a plurality of pixel values of the image of the second subject captured through the filter array,
- the neural network generates the plurality of output values using the plurality of input values and the determined plurality of weight values and without using the restoration matrix.
- the technology of the present disclosure is useful, for example, for cameras and measuring instruments that acquire multi-wavelength or high-resolution images.
- the technology of the present disclosure can also be applied to, for example, sensing for living organisms, medical care, and beauty, food foreign matter/residual pesticide inspection systems, remote sensing systems, and in-vehicle sensing systems.
- Imaging device 110 Filter array 140 Optical system 160 Image sensor 200 Signal processing device 210 Processor 250 Memory 300 Display device 400 Conveyor 500 Terminal device 510 Input device 520 Display device 1000 Imaging system
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Multimedia (AREA)
- Human Computer Interaction (AREA)
- Spectroscopy & Molecular Physics (AREA)
- Software Systems (AREA)
- Artificial Intelligence (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Medical Informatics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Signal Processing (AREA)
- Image Processing (AREA)
Abstract
Description
情報量の少ない圧縮画像から、より多くの情報を含む復元画像を復元する処理には、様々なアルゴリズムが適用され得る。例えば、前述の圧縮センシング技術に基づく種々のアルゴリズム、または深層学習などの機械学習に基づく種々のアルゴリズムが利用され得る。個々のアルゴリズムは固有の特徴を有する。例えば、あるアルゴリズムを用いた場合、高精度の復元が可能であるが、演算量が多く、復元処理に長い時間を要することがある。一方、他のアルゴリズムを用いた場合、短い時間で復元が可能であるが、復元精度が劣る場合がある。
(S11)被写体の圧縮された情報を含む圧縮画像を取得する。
(S12)圧縮画像から復元目標画像を生成するための第1復元処理に用いられる第1パラメータ群および復元行列を取得する。
(S13)圧縮画像、第1パラメータ群、および復元行列に基づいて、復元目標画像を生成する。
(S14)圧縮画像から復元画像を生成するための第2復元処理に用いられる第2パラメータ群を取得する。
(S15)圧縮画像、第2パラメータ群、および復元行列に基づいて、復元画像を生成する。
(S16)復元目標画像および復元画像に基づいて、第2パラメータ群を補正する。
次に、本開示の例示的な実施形態において用いられ得る撮像システムの構成例を説明する。
上記のような撮像システムを利用して、例えば搬送装置によって搬送される対象物をハイパースペクトル画像に基づいて検査または分析するシステムを構築することができる。そのようなシステムにおいて、リアルタイムの検査または分析を実現するためには、圧縮画像から復元画像を生成するための復元処理を短時間で実行することが求められる。しかし、圧縮センシングに基づく既存のアルゴリズムでは、復元性能は高いものの、計算量が多く、要求される短い時間内で処理を完了することができないことがあった。一方、既存のアルゴリズムの中には、高速な処理が可能であるものの、設定すべきパラメータが多く、パラメータの最適化が困難または煩雑であるアルゴリズムも存在する。
・撮像装置100によって生成された圧縮画像を取得する。
・第1パラメータ群および復元行列(前述の行列Hに相当)をメモリ250から取得する。
・圧縮画像、第1パラメータ群、および復元行列に基づいて、復元目標画像を生成する。
・第2パラメータ群をメモリ250から取得する。
・圧縮画像、第2パラメータ群、および復元行列に基づいて、復元画像を生成する。
・復元目標画像および復元画像に基づいて、第2パラメータ群を補正する。
これにより、領域Bの誤差は評価値にそれほど影響せず、領域Aの誤差が評価値に大きく影響することになる。係数aは、係数bよりも大きい値、例えば係数bの1.5倍、2倍、またはそれ以上の値に設定され得る。このような処理により、重要な被写体が存在する領域における復元精度を優先した復元が実行される。
以上の実施の形態の記載により、以下の技術が開示される。
コンピュータを用いて実行される信号処理方法であって、
被写体の圧縮された情報を含む圧縮画像を取得することと、
前記圧縮画像から復元目標画像を生成するための第1復元処理に用いられる第1パラメータ群および復元行列を取得することと、
前記圧縮画像、前記第1パラメータ群、および前記復元行列に基づいて、前記復元目標画像を生成することと、
前記圧縮画像から復元画像を生成するための第2復元処理に用いられる第2パラメータ群を取得することと、
前記圧縮画像、前記第2パラメータ群、および前記復元行列に基づいて、前記復元画像を生成することと、
前記復元目標画像および前記復元画像に基づいて、前記第2パラメータ群を補正することと、
を含む方法。
前記第2パラメータ群は、前記第1パラメータ群よりも多くのパラメータを含む、技術1に記載の方法。
前記第2パラメータ群を補正することは、前記復元画像が前記復元目標画像に近づくように、前記第2パラメータ群に含まれる1つ以上のパラメータに対応する1つ以上の値を補正することを含む、技術1または2に記載の方法。
前記第2パラメータ群を補正することは、
前記目標復元画像と前記復元画像との誤差評価値を求めることと、
前記誤差評価値を最小化するように前記第2パラメータ群に含まれる1つ以上のパラメータに対応する1つ以上の値を補正することと、
を含む、技術3に記載の方法。
前記第1復元処理は、第1アルゴリズムによって実行され、
前記第2復元処理は、第2アルゴリズムによって実行され、
前記第2アルゴリズムは、前記第1アルゴリズムよりも演算負荷の小さいアルゴリズムである、
技術1から4のいずれかに記載の方法。
前記第1復元処理は、機械学習を通じて学習される学習済みモデルに基づく処理を含まず、
前記第2復元処理は、機械学習を通じて学習される学習済みモデルに基づく処理を含む、
技術5に記載の方法。
前記第1復元処理は、前記圧縮画像および前記復元行列に基づく評価関数を最小化または最大化する反復演算を含む、技術6に記載の方法。
前記圧縮画像は、前記被写体のスペクトル情報が符号化された画像であり、
前記復元目標画像および前記復元画像の各々は、複数の波長バンドに対応する複数の画像の情報を含む、
技術1から7のいずれかに記載の方法。
前記復元目標画像および前記復元画像に基づいて前記第2パラメータ群を補正することと、補正後の前記第2パラメータ群を用いて前記復元画像を生成することとを、複数回繰り返すことによって最終的な前記第2パラメータ群を決定することを含む、技術1から8のいずれかに記載の方法。
ユーザが前記第1パラメータ群を入力するためのグラフィカルユーザインターフェース(GUI)を表示装置に表示させることをさらに含む、技術1から9のいずれかに記載の方法。
前記復元目標画像および前記復元画像に基づいて前記第2パラメータ群を補正することと、補正後の前記第2パラメータ群を用いて前記復元画像を生成することとを、終了条件を満たさない限り所定回数繰り返すことと、
前記所定回数の繰り返しが行われた後の前記復元画像と前記復元目標画像との誤差評価値を計算することと、
前記誤差評価値が閾値よりも大きい場合に、前記第1パラメータ群の再入力、前記所定回数の変更、および前記終了条件の変更の少なくとも1つを前記ユーザに促すGUIを前記表示装置に表示させることをさらに含む、
技術10に記載の方法。
前記誤差評価値を計算することは、
前記復元目標画像における第1領域を抽出することと、
前記復元画像において前記第1領域に対応する第2領域を抽出することと、
前記第1領域と前記第2領域との差異に基づいて前記誤差評価値を決定することと、
を含む、技術11に記載の方法。
信号処理装置であって、
1つ以上のプロセッサと、
前記1つ以上のプロセッサによって実行されるコンピュータプログラムを格納したメモリと、
を備え、
前記コンピュータプログラムは、前記1つ以上のプロセッサに、
被写体の圧縮された情報を含む圧縮画像を取得することと、
前記圧縮画像から復元目標画像を生成するための第1復元処理に用いられる第1パラメータ群および復元行列を取得することと、
前記圧縮画像、前記第1パラメータ群、および前記復元行列に基づいて、前記復元目標画像を生成することと、
前記圧縮画像から復元画像を生成するための第2復元処理に用いられる第2パラメータ群を取得することと、
前記圧縮画像、前記第2パラメータ群、および前記復元行列に基づいて、前記復元画像を生成することと、
前記復元目標画像および前記復元画像に基づいて、前記第2パラメータ群を補正することと、
を実行させる、信号処理装置。
本開示の実施の形態の変形例は下記に示すようなものであってもよい。
前記方法は、コンピュータが
(a)4つ以上のフィルタを含むフィルタアレイを通して撮像された被写体の第1画像を受け取り、前記4つ以上のフィルタは4つの以上の分光スペクトルに1対1対応し、前記4つの以上の分光スペクトルは互いに異なり、
(b)前記フィルタアレイに対応する復元行列と前記第1画像の複数の画素値に基づいて、第1波長域に対応する第1復元目標画像の複数の第1画素値、~、第n波長域に対応する第n復元目標画像の複数の第n画素値を決定し、
(c)ニューラルネットワークの複数の重み値を、前記第1画像の前記複数の画素値、前記複数の第1画素値、~、前記複数の第n画素値に基づいて決定し、
前記複数の重み値のそれぞれは直接接続された2つのユニットに対応し、前記ニューラルネットワークは前記2つのユニットを含み、
前記複数の第1画素値、~、前記複数の第n画素値は、前記第1画像の前記複数の画素値が前記ニューラルネットワークが入力された場合に、前記ニューラルネットワークが出力する複数の出力値に対する教師データとして用いられ、
(d)前記ニューラルネットワークへの複数の入力値に対する前記ニューラルネットワークから複数の出力値に基づいて、第2被写体の前記第1波長域に対応する画像、~、または、前記第2被写体の前記第n波長域に対応する画像を生成し、
前記複数の入力値は前記フィルタアレイを通して撮像された前記第2被写体の画像の複数の画素値であり、
前記ニューラルネットワークが前記複数の入力値と前記決定された複数の重み値を用いて、かつ、前記復元行列を用いず、前記複数の出力値を生成する。
110 フィルタアレイ
140 光学系
160 イメージセンサ
200 信号処理装置
210 プロセッサ
250 メモリ
300 表示装置
400 コンベア
500 端末装置
510 入力装置
520 表示装置
1000 撮像システム
Claims (13)
- コンピュータを用いて実行される信号処理方法であって、
被写体の圧縮された情報を含む圧縮画像を取得することと、
前記圧縮画像から復元目標画像を生成するための第1復元処理に用いられる第1パラメータ群および復元行列を取得することと、
前記圧縮画像、前記第1パラメータ群、および前記復元行列に基づいて、前記復元目標画像を生成することと、
前記圧縮画像から復元画像を生成するための第2復元処理に用いられる第2パラメータ群を取得することと、
前記圧縮画像、前記第2パラメータ群、および前記復元行列に基づいて、前記復元画像を生成することと、
前記復元目標画像および前記復元画像に基づいて、前記第2パラメータ群を補正することと、
を含む方法。 - 前記第2パラメータ群は、前記第1パラメータ群よりも多くのパラメータを含む、請求項1に記載の方法。
- 前記第2パラメータ群を補正することは、
前記復元画像が前記復元目標画像に近づくように、前記第2パラメータ群に含まれる1つ以上のパラメータに対応する1つ以上の値を補正することを含む、請求項1または2に記載の方法。 - 前記第2パラメータ群を補正することは、
前記目標復元画像と前記復元画像との誤差評価値を求めることと、
前記誤差評価値を最小化するように前記第2パラメータ群に含まれる1つ以上のパラメータに対応する1つ以上の値を補正することと、
を含む、請求項3に記載の方法。 - 前記第1復元処理は、第1アルゴリズムによって実行され、
前記第2復元処理は、第2アルゴリズムによって実行され、
前記第2アルゴリズムは、前記第1アルゴリズムよりも演算負荷の小さいアルゴリズムである、
請求項1または2に記載の方法。 - 前記第1復元処理は、機械学習を通じて学習される学習済みモデルに基づく処理を含まず、
前記第2復元処理は、機械学習を通じて学習される学習済みモデルに基づく処理を含む、
請求項5に記載の方法。 - 前記第1復元処理は、前記圧縮画像および前記復元行列に基づく評価関数を最小化または最大化する反復演算を含む、請求項6に記載の方法。
- 前記圧縮画像は、前記被写体のスペクトル情報が符号化された画像であり、
前記復元目標画像および前記復元画像の各々は、複数の波長バンドに対応する複数の画像の情報を含む、
請求項1または2に記載の方法。 - 前記復元目標画像および前記復元画像に基づいて前記第2パラメータ群を補正することと、補正後の前記第2パラメータ群を用いて前記復元画像を生成することとを、複数回繰り返すことによって最終的な前記第2パラメータ群を決定することを含む、請求項1または2に記載の方法。
- ユーザが前記第1パラメータ群を入力するためのグラフィカルユーザインターフェース(GUI)を表示装置に表示させることをさらに含む、請求項1または2に記載の方法。
- 前記復元目標画像および前記復元画像に基づいて前記第2パラメータ群を補正することと、補正後の前記第2パラメータ群を用いて前記復元画像を生成することとを、終了条件を満たさない限り所定回数繰り返すことと、
前記所定回数の繰り返しが行われた後の前記復元画像と前記復元目標画像との誤差評価値を計算することと、
前記誤差評価値が閾値よりも大きい場合に、前記第1パラメータ群の再入力、前記所定回数の変更、および前記終了条件の変更の少なくとも1つを前記ユーザに促すGUIを前記表示装置に表示させることをさらに含む、
請求項10に記載の方法。 - 前記誤差評価値を計算することは、
前記復元目標画像における第1領域を抽出することと、
前記復元画像において前記第1領域に対応する第2領域を抽出することと、
前記第1領域と前記第2領域との差異に基づいて前記誤差評価値を決定することと、
を含む、請求項11に記載の方法。 - 信号処理装置であって、
1つ以上のプロセッサと、
前記1つ以上のプロセッサによって実行されるコンピュータプログラムを格納したメモリと、
を備え、
前記コンピュータプログラムは、前記1つ以上のプロセッサに、
被写体の圧縮された情報を含む圧縮画像を取得することと、
前記圧縮画像から復元目標画像を生成するための第1復元処理に用いられる第1パラメータ群および復元行列を取得することと、
前記圧縮画像、前記第1パラメータ群、および前記復元行列に基づいて、前記復元目標画像を生成することと、
前記圧縮画像から復元画像を生成するための第2復元処理に用いられる第2パラメータ群を取得することと、
前記圧縮画像、前記第2パラメータ群、および前記復元行列に基づいて、前記復元画像を生成することと、
前記復元目標画像および前記復元画像に基づいて、前記第2パラメータ群を補正することと、
を実行させる、信号処理装置。
Priority Applications (4)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202380057930.8A CN119654651A (zh) | 2022-08-25 | 2023-07-31 | 信号处理方法及信号处理装置 |
| JP2024542703A JPWO2024043009A1 (ja) | 2022-08-25 | 2023-07-31 | |
| EP23857124.4A EP4579576A4 (en) | 2022-08-25 | 2023-07-31 | SIGNAL PROCESSING METHOD AND SIGNAL PROCESSING DEVICE |
| US19/038,677 US20250173055A1 (en) | 2022-08-25 | 2025-01-27 | Signal processing method and signal processing apparatus |
Applications Claiming Priority (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| JP2022-134391 | 2022-08-25 | ||
| JP2022134391 | 2022-08-25 |
Related Child Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| US19/038,677 Continuation US20250173055A1 (en) | 2022-08-25 | 2025-01-27 | Signal processing method and signal processing apparatus |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| WO2024043009A1 true WO2024043009A1 (ja) | 2024-02-29 |
Family
ID=90013060
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| PCT/JP2023/027993 Ceased WO2024043009A1 (ja) | 2022-08-25 | 2023-07-31 | 信号処理方法および信号処理装置 |
Country Status (5)
| Country | Link |
|---|---|
| US (1) | US20250173055A1 (ja) |
| EP (1) | EP4579576A4 (ja) |
| JP (1) | JPWO2024043009A1 (ja) |
| CN (1) | CN119654651A (ja) |
| WO (1) | WO2024043009A1 (ja) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP7687748B1 (ja) * | 2024-10-21 | 2025-06-03 | 国立研究開発法人農業・食品産業技術総合研究機構 | 情報処理装置、情報処理方法、及びプログラム |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2016156801A (ja) * | 2014-11-19 | 2016-09-01 | パナソニックIpマネジメント株式会社 | 撮像装置および分光システム |
| JP2017208641A (ja) | 2016-05-17 | 2017-11-24 | キヤノン株式会社 | 圧縮センシングを用いた撮像装置、撮像方法および撮像プログラム |
| US20190340497A1 (en) | 2016-12-09 | 2019-11-07 | William Marsh Rice University | Signal Recovery Via Deep Convolutional Networks |
| WO2021234797A1 (ja) * | 2020-05-18 | 2021-11-25 | 日本電信電話株式会社 | 学習方法、高分解能化方法、学習装置及びコンピュータープログラム |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2022075094A1 (ja) * | 2020-10-05 | 2022-04-14 | パナソニックIpマネジメント株式会社 | 処理装置、システム、方法、コンピュータプログラム |
-
2023
- 2023-07-31 JP JP2024542703A patent/JPWO2024043009A1/ja active Pending
- 2023-07-31 EP EP23857124.4A patent/EP4579576A4/en active Pending
- 2023-07-31 CN CN202380057930.8A patent/CN119654651A/zh active Pending
- 2023-07-31 WO PCT/JP2023/027993 patent/WO2024043009A1/ja not_active Ceased
-
2025
- 2025-01-27 US US19/038,677 patent/US20250173055A1/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2016156801A (ja) * | 2014-11-19 | 2016-09-01 | パナソニックIpマネジメント株式会社 | 撮像装置および分光システム |
| US9599511B2 (en) | 2014-11-19 | 2017-03-21 | Panasonic Intellectual Property Management Co., Ltd. | Imaging apparatus comprising coding element and spectroscopic system comprising the imaging apparatus |
| JP2017208641A (ja) | 2016-05-17 | 2017-11-24 | キヤノン株式会社 | 圧縮センシングを用いた撮像装置、撮像方法および撮像プログラム |
| US20190340497A1 (en) | 2016-12-09 | 2019-11-07 | William Marsh Rice University | Signal Recovery Via Deep Convolutional Networks |
| WO2021234797A1 (ja) * | 2020-05-18 | 2021-11-25 | 日本電信電話株式会社 | 学習方法、高分解能化方法、学習装置及びコンピュータープログラム |
Non-Patent Citations (3)
| Title |
|---|
| See also references of EP4579576A4 |
| WANG LIZHI; SUN CHEN; FU YING; KIM MIN H.; HUANG HUA: "Hyperspectral Image Reconstruction Using a Deep Spatial-Spectral Prior", 2019 IEEE/CVF CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION (CVPR), IEEE, 15 June 2019 (2019-06-15), pages 8024 - 8033, XP033687126, DOI: 10.1109/CVPR.2019.00822 * |
| YOSHIDA MICHITAKA, NAGAHARA HAJIME, TORII AKIHIKO, OKUTOMI MASATOSHI, TANIGUCHI RIN-ICHIRO: "Reconstruction of compressed video sensing by Deep Learning", IPSJ SIG TECHNICAL REPORT. CVIM-208, vol. 2017, no. 19, 15 September 2017 (2017-09-15), XP093142965 * |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP7687748B1 (ja) * | 2024-10-21 | 2025-06-03 | 国立研究開発法人農業・食品産業技術総合研究機構 | 情報処理装置、情報処理方法、及びプログラム |
Also Published As
| Publication number | Publication date |
|---|---|
| JPWO2024043009A1 (ja) | 2024-02-29 |
| EP4579576A4 (en) | 2025-12-10 |
| CN119654651A (zh) | 2025-03-18 |
| EP4579576A1 (en) | 2025-07-02 |
| US20250173055A1 (en) | 2025-05-29 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| JP7720532B2 (ja) | 信号処理方法、信号処理装置、および撮像システム | |
| JP7457952B2 (ja) | 光検出装置、光検出システム、およびフィルタアレイ | |
| US11843876B2 (en) | Optical filter array, photodetection device, and photodetection system | |
| JP7653612B2 (ja) | 光検出装置、光検出システム、およびフィルタアレイ | |
| JP7668452B2 (ja) | 信号処理方法、信号処理装置、および撮像システム | |
| JP7142251B2 (ja) | フィルタアレイおよび光検出システム | |
| JP7829156B2 (ja) | 画像処理装置、撮像システム、および復元画像の誤差を推定する方法 | |
| US20250338030A1 (en) | Filter array, photodetector, and photodetection system | |
| WO2021171905A1 (ja) | 撮像装置 | |
| JP7811739B2 (ja) | 検査対象に含まれる異物を検出する方法および装置 | |
| US20250173055A1 (en) | Signal processing method and signal processing apparatus | |
| US20240311971A1 (en) | Signal processing method, non-volatile computer-readable recording medium, and system | |
| JP7122636B2 (ja) | フィルタアレイおよび光検出システム | |
| WO2026088780A1 (ja) | 情報処理方法、処理装置、およびコンピュータプログラム | |
| US12604073B2 (en) | Device and filter array used in system for generating spectral images corresponding to N wavelength bands where N is an integer greater than or equal to four | |
| JP2025071703A (ja) | 画像を生成するシステムに用いられる装置およびフィルタアレイ | |
| JP2024020922A (ja) | 復元画像の評価方法および撮像システム、 | |
| WO2024195499A1 (ja) | 撮像システム、行列データ、および行列データの生成方法 | |
| WO2025084081A1 (ja) | 撮像システムおよびそれに用いられる方法 | |
| WO2025192316A1 (ja) | 撮像装置、撮像システム、および撮像方法 | |
| WO2024053302A1 (ja) | 情報処理方法および撮像システム | |
| WO2024203282A1 (ja) | 撮像システムおよび方法 | |
| WO2025192485A1 (ja) | 撮像システムおよび撮像方法 | |
| WO2024142683A1 (ja) | 光検出装置、光検出システム、およびフィルタアレイ | |
| Stockmans et al. | Di k, C. an, & Keller, CU (2023) |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| 121 | Ep: the epo has been informed by wipo that ep was designated in this application |
Ref document number: 23857124 Country of ref document: EP Kind code of ref document: A1 |
|
| WWE | Wipo information: entry into national phase |
Ref document number: 2024542703 Country of ref document: JP |
|
| WWE | Wipo information: entry into national phase |
Ref document number: 202380057930.8 Country of ref document: CN |
|
| WWP | Wipo information: published in national office |
Ref document number: 202380057930.8 Country of ref document: CN |
|
| WWE | Wipo information: entry into national phase |
Ref document number: 2023857124 Country of ref document: EP |
|
| NENP | Non-entry into the national phase |
Ref country code: DE |
|
| ENP | Entry into the national phase |
Ref document number: 2023857124 Country of ref document: EP Effective date: 20250325 |
|
| WWP | Wipo information: published in national office |
Ref document number: 2023857124 Country of ref document: EP |