JPH0877351A - Image processing method and apparatus - Google Patents

Abstract

(57) [Abstract] [Purpose] To provide an image processing technique capable of processing a read image at high speed even if the image is deformed. [Structure] Image data of an object is read by an image reading unit 1, image data is accumulated in an image memory 2, and data of a deformed image is converted into regular image data 6 based on a calculation result of a deformation amount of the accumulated image data. The data is transferred to the pattern matching processing unit 9 while being converted. At the same time as the transfer, the pattern matching processing unit 9 compares the regular image data in the dictionary memory 10 with the transferred data to perform the pattern matching processing.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a technique of image processing effectively applied to recognition of character information, positioning of parts, visual inspection and the like.

[0002]

2. Description of the Related Art For example, in order to facilitate searching and editing of a document, character information of a document read by an image reading device such as a video camera or an image scanner is image-processed and converted into a character code. I'm trying to convert. In such image processing technology, the image data of the read character is identified by matching the image data of the read character with the character pattern registered in the dictionary in advance corresponding to the character code. I am trying to do it.

When a document is read by a video camera or the like, if the posture of the character is tilted or curved, the character cannot be discriminated. Although it is necessary to devise a system and an illumination system, the manufacturing cost of an optical system etc. will increase.

Therefore, when a deformed image such as a tilted image or a curved image is captured when the image is read, the pattern is converted after the image is converted into a regular image by using software calculation or DSP. The matching is done.

For example, when an image is read with an inclination, the inclination angle θ is obtained from the reference position image of the input image by software calculation, and the normal position is obtained based on the inclination angle θ, issued by Kyoritsu Publishing Co., Ltd. "Image processing (2nd edition)" 1989
Published October 15, 2015, by Hiroshi Ozaki and Keiji Taniguchi, P105-P
Techniques for affine transformation have been developed, as described at 106.

In order to accurately recognize a character read by tilting, the tilt angle is obtained, and the tilted image data is corrected to generate a normal image and then compared with the character pattern registered in the dictionary. There are a method of performing pattern matching and a method of performing pattern matching processing by rotating a character pattern registered in advance by a required inclination angle. Further, a character pattern is registered in the dictionary corresponding to each inclination angle, a character pattern corresponding to the obtained inclination angle is extracted from the dictionary, and the read image data and the character pattern are subjected to pattern matching processing. Schemes are being developed.

[0007]

However, when the read image is deformed, the image may be converted into a regular image or the dictionary pattern may be rotated to perform pattern matching. It takes more than twice as long as usual to process, and it takes time to recognize characters. In addition, registering a dictionary pattern for each inclination angle requires a memory capacity for storing a huge amount of data for that purpose.

An object of the present invention is to provide an image processing technique capable of processing a read image at high speed even if the image is deformed.

The above and other objects and novel features of the present invention will be apparent from the description of this specification and the accompanying drawings.

[0010]

Of the inventions disclosed in the present application, a representative one will be briefly described below.
It is as follows.

That is, according to the image processing method of the present invention, the image data of the object is read by the image reading unit, the image data is accumulated in the image memory, and the deformed image is calculated based on the calculation result of the deformation amount of the accumulated image data. Is converted into the data of the regular image and transferred to the pattern matching processing unit, and the pattern matching processing unit compares the normal image data in the comparison image memory with the transferred data and performs the pattern matching processing at the same time. And

Further, the image processing apparatus of the present invention has an address generator for converting the image data stored in the image memory into a normal image based on the image calculation parameter of the deformation amount, and the address generator converts the image data into a normal image. It is characterized in that the transfer data transferred from the image memory is compared with the regular image data in the comparison image memory and the image of the transfer data is subjected to pattern matching processing.

Further, the image processing apparatus of the present invention includes an address generator for designating an address of data transferred from the image data stored in the image memory on the basis of the calculation parameter of the tilt angle calculated by the calculation unit, and an image generator. An image cropping circuit for converting the transfer data transferred from the memory in the order set by the address generator into a two-dimensional image, and comparing the two-dimensional image from the image cropping circuit with the normal image data in the comparison image memory. The feature is that pattern matching processing for identifying an image is performed. Then, the comparison image memory is a dictionary memory in which a plurality of character patterns are registered together with respective character codes,
You may make it convert the character data of the target object read by the image reading part into a character code.

[0014]

When the image data of the object read by the image reading unit and stored in the image memory is transferred to the pattern matching processing unit, the image data of the object is transferred back to the normal image according to the deformation amount of the image data of the object. It Then, pattern matching processing is performed on the regular image data and the transfer image data registered in the comparison image memory. Therefore, the image memory that accumulates the image data read by the image reading unit as it is also functions as a virtual memory that performs address rearrangement in combination with the address generator. It is possible to quickly perform image processing without converting the image data stored in the image memory according to the deformation amount and rearranging the image data in another memory or the like.

[0015]

1 is a diagram showing an image processing apparatus according to an embodiment of the present invention. In the case shown, this image processing apparatus is used for recognizing characters written in a document or the like. There is.

The image reading section 1 is composed of a video camera, an image scanner, or the like. The video camera converts an original image formed on the photoelectric conversion element by a lens into an electric signal in time series and outputs a video signal. Further, the image scanner has a line sensor in which CCD elements are linearly arranged. By scanning and moving the CCD with respect to the original, one image of the original is read and the image of the original is converted into an electronic signal. Convert.

The image reading section 1 thus formed by the video camera has an amplifier for amplifying the electric signal converted by the element, an A / D converter for converting it into a digital signal, and the like. The unit 1 is connected to an image memory 2 for storing A / D converted image information, that is, a frame memory. When the image information is stored in the image memory 2, the image memory 2 is accessed via the address bus from the CPU 3.

FIG. 1 shows a state in which the image of the character "F" read by the image reading section 1 is stored in the image memory 2 in a state of being tilted by a tilt angle θ with respect to the regular image. This inclination angle θ is obtained by the CPU 3 taking in the image information stored in the image memory 2 and calculating it by software.

The CPU 3 sends a calculation parameter for image conversion to the address generator 4 through the data bus 5 on the basis of the tilt angle θ calculated and calculated. The address generator 4 is composed of an image conversion LSI that is programmable and can generate an arbitrary address. To convert a tilted image into a regular image, a CPU is used.
An X coordinate address 4x and a Y coordinate address 4y are generated in the image memory 2 based on the calculation parameter from 3 and the calculation parameter for image conversion is sent to the address generator 4. In this way, by controlling the transfer of the image data from the image memory 2 by the address generator 4, the image memory 2 can be made to function as a virtual memory.

Based on the operation parameter, the normal image data 6 designated by the X, Y coordinate addresses 4x, 4y to become a normal image is serially transferred to the image cutting circuit 7 from the image memory 2. This image cutout circuit 7
Is a circuit for converting serially transferred image data into a two-dimensional image, and is composed of a shift register.

FIG. 2 is a diagram conceptually showing a procedure for transferring the image data accumulated in the image memory 2 in a tilted posture to the image cutting circuit 7 as the normal image data 6. As shown in FIG. 2, when the image data is transferred from the image memory 2, the image data is scanned in the direction indicated by the arrow based on the calculation parameter from the CPU 3, and the image cutting circuit 7 serially outputs the image data. Image data is transferred.

From the image cut-out circuit 7, the regular image data 8 is transferred in parallel as two-dimensional image data to the pattern matching processing section 9 composed of an LSI. On the other hand, the pattern matching processing unit 9 is connected to the dictionary memory 10, and the character pattern is registered in the dictionary memory 10 in a normal posture. Therefore, the data of the transferred image, which is transferred to the pattern matching processing 9 and becomes a normal image, is compared with the dictionary pattern sent from the dictionary memory 10 to the pattern matching processing unit 9 via the dictionary pattern data bus 11 as it is, and the pattern matching is performed. Processing is performed.

If the regular image data 8 is, for example, the data of the character "F" by this pattern matching processing, the character is recognized by comparing with the dictionary pattern. As a result, this pattern matching processing unit 9
Is converted into a character code and the signal is output to an output unit (not shown).

As shown in FIG. 2, if the image data stored in the image memory 2 in an inclined state is transferred as it is, the pattern matching processing section 9 cannot perform proper processing. , The transfer order is scanned in the image memory 2 based on the designation of the coordinate addresses 4x and 4y, so that even if the image data is accumulated in the image memory 2 in a tilted state, the pattern matching processing unit 9 On the other hand, it will be transferred as image data of a normal posture.

Therefore, it is not necessary to convert the posture of the image data stored in the image memory 2 to another memory, convert the image data into a normal image data, and store the data again.
It is possible to quickly perform character recognition processing without requiring an extra memory. Further, it is not necessary to register the character pattern for each inclination angle in the dictionary memory 10, and it is not necessary to increase the storage capacity of the dictionary memory 10.

Although the invention made by the present inventor has been specifically described based on the embodiments, the present invention is not limited to the embodiments and various modifications can be made without departing from the scope of the invention. Needless to say.

For example, in the illustrated embodiment, the image processing technique of the present invention is applied for recognizing characters, but the present invention can also be applied for positioning, measuring and visual inspection of parts. In that case, the shapes to be compared such as the position and appearance of parts are registered in advance in the comparison image memory, the image of the object is read and pattern matching processing is performed, and the position and appearance of parts are appropriate. Can be detected.

In the above description, the case where the invention made by the present inventor is mainly applied to the technology of character recognition, which is the field of use of the invention, has been described, but the invention is not limited to this. It is also applicable to the labeling processing technique.

[0029]

Advantageous effects obtained by typical ones of the inventions disclosed in the present application will be briefly described.
It is as follows.

(1). Since the pattern matching processing can be performed while converting the deformed image into the regular image, the image processing can be performed at high speed.

(2) It is possible to simplify the positioning accuracy of the object when the object is read by the image reading section, and it becomes unnecessary to use expensive optical and illumination systems.
The cost performance can be improved.

(3) Therefore, it is possible to obtain an image processing device that can perform high-speed processing at low cost.

[Brief description of drawings]

FIG. 1 is a block diagram showing the structure of an image processing apparatus that is an embodiment of the present invention.

FIG. 2 is a schematic diagram showing a basic concept of image conversion.

[Explanation of symbols]

 1 Image Reading Unit 2 Image Memory 3 CPU (Calculation Unit) 4 Address Generator 4x X Coordinate Address 4y Y Coordinate Address 5 Data Bus 6 Regular Image Data 7 Image Cutout Circuit 8 Regular Image Data 9 Pattern Matching Processing Unit 10 Dictionary Memory (Comparison Image) Memory) 11 Dictionary pattern data bus

Claims (4)

[Claims] 1. A step of reading image data of an object by an image reading unit and accumulating the image data in an image memory,
A step of transferring the data of the deformed image to the data of the regular image while converting the data of the deformed image to the data of the regular image based on the accumulated calculation result of the deformation amount of the image data; And a pattern matching process by comparing the regular image data with the transfer data. 2. An image reading unit for reading image data of an object, an image memory for storing the image data read by the image reading unit, and an operation for calculating a deformation amount of the image data stored in the image memory. Section, an address generator that converts the image data accumulated in the image memory into a normal image based on the image calculation parameter of the deformation amount calculated by the calculation section, and a comparison image memory in which the normal image data is registered. A pattern matching processing unit for identifying an image of the transfer data by comparing the transfer data converted into the normal image by the address generator and transferred from the image memory with the normal image data of the comparison image memory. A characteristic image processing device. 3. An image reading unit for reading image data of an object, an image memory for storing the image data read by the image reading unit, and a calculation for calculating an inclination angle of the image data stored in the image memory. Section, an address generator for designating the address of the data transferred from the image data stored in the image memory based on the calculation parameter of the tilt angle calculated by the calculation section, and the address generator from the image memory by the address generator. An image cutout circuit for converting transfer data transferred in a set order into a two-dimensional image, a comparison image memory in which regular image data is registered, a two-dimensional image from the image cutout circuit, and a regular image of the comparison image memory. An image characterized by having a pattern matching processing section for comparing the image data and identifying the image Processing equipment. 4. The comparison image memory is a dictionary memory in which a plurality of character patterns are registered together with respective character codes, and the character data of the object read by the image reading unit is converted into a character code. The image processing apparatus according to claim 2 or 3, characterized in that