JPH03214272A - Selective smoothing filter mechanism using edge information - Google Patents

Abstract

(57) [Summary] This bulletin contains application data before electronic filing, so abstract data is not recorded.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to an image processing device, and particularly to a mechanism for performing 1-to-1 smoothing processing on an image.

[Prior Art] In the field of image processing, image information is generally processed digitally. One such digital image processing technique is called smoothing processing. This smoothing process is a process that mainly removes notch components, random noise, etc. contained in image information.
It has a function equivalent to a low-pass filter.

Such smoothing processing not only removes high-frequency components such as random noise, but also subtracts a constant multiple of the smoothed original image, thereby improving the sharpness of the image and improving the boundary area (edge area) of the image. It is also used in processes such as emphasizing.

Image smoothing is generally defined as a process that outputs the output value of a pixel P as a function h (f) of the density (or brightness) of this pixel P and input pixels (original pixels) in its vicinity. be done. Here, f indicates the density value (or brightness) of the pixel P and the input image pixels in its vicinity. There are various processing techniques for such smoothing processing, such as uniform weight smoothing processing, first-order difference smoothing processing, and smoothing processing using the average value or median value of density values in neighboring regions. There is.

Next, conventional smoothing processing will be briefly explained with reference to FIG. FIG. 5(a) exemplarily shows the luminance value of each pixel in a part of the screen. In FIG. 5(a), each square represents a pixel, and the value within each square represents the brightness of the pixel. This display method is assumed to be the same in the following explanation.

Now, for each pixel P (i. Output value P after processing
Let us consider the case where the process of outputting as '(i,j), ie, the process of P'(i.j)=iP(i,j)+P(i,j+1)l/2, is performed. Here, for convenience of explanation, P
(i, j) indicates a pixel located at row l and column J in the pixel matrix of the screen, and also indicates its luminance value. The luminance value of each pixel after performing the above-mentioned smoothing process on the image shown in FIG. 5(a) is shown in FIG. 5(b).
).

In FIG. 5(b), of the pixel matrix of 4 rows and 7 columns, the luminance values of the pixels in the 1st column to the 3rd column are "1", and the luminance values of the pixels in the 5th column
The brightness values of the pixels in the columns to the seventh column are "11", and the brightness values of the pixels in the fourth column to
It can be seen that the luminance value of the pixels in the column is 6 degrees.

Therefore, as is clear from comparing Figures 5(a) and 5(b), the change in brightness between valley pixels becomes gradual, and the image is smoothed. As can be seen, in FIG. 5Cb), it can be seen that the fourth column is the pixel column that constitutes the boundary area (Etsu 2 area) of the image.

Problem R to be Solved by the Invention] As described above, in the conventional smoothing process, pixels in the vicinity of each pixel are used to smooth the image. At this time, no consideration is given to whether the neighboring area includes an edge area or whether the pixel to be processed is included in an edge area, and the same smoothing process is performed on all pixels. It was.

Therefore, as shown in Figure 5(b), although the entire image is smoothed and the edge area is preserved in the image after the smoothing process, this edge area is also smoothed, so the vicinity of the edge area The problem arises that the difference in brightness between the edges becomes smaller, and the negative effect of smoothing occurs near this edge region, that is, the edge of the image deteriorates, and the boundary region becomes unclear.

SUMMARY OF THE INVENTION Therefore, it is an object of the present invention to provide a smoothing filter mechanism that eliminates the drawbacks of conventional smoothing processing techniques and can smooth images while keeping edge regions sharp.

[Means for Solving the Problems] The smoothing filter mechanism according to the present invention provides a smoothing filter mechanism for a certain pixel P depending on whether or not a pixel includes a pixel constituting an edge region in its neighboring region. Either a smoothing process is performed or the pixel P is passed through as is.

That is, the smoothing filter mechanism according to the present invention includes a means for deriving a pixel string constituting an image to be processed, a means for performing a predetermined smoothing process on the pixel string from the deriving means, and a means for deriving a pixel string from the deriving means. a means for detecting an edge region included in an image from a pixel column, and whether smoothing processing in the smoothing processing means was performed using pixels included in the edge region of the image in response to the output of the detection means; and means for outputting either the pixel from the deriving means or the pixel from the smoothing processing means as a processed pixel in response to the output of the determining means.

The output means passes the pixel P from the derivation means when the determination means determines that a certain pixel P is subjected to smoothing processing using a pixel included in an edge region, and otherwise passes the pixel P from the derivation means. Pixel data corresponding to pixel P from the smoothing processing means is passed.

[Operation] In the above configuration, pixel data before processing is output for pixels that are included in the edge area and are subjected to smoothing processing, while for pixels that are not included in the edge region, smoothing processing is performed. Processed pixel data is output, so
A smoothed image is obtained in which the edge area of the original image and the image in its vicinity are not smoothed, but the Enon area of the original image is preserved, and a smoothed image with clear border areas of the image is obtained.

[Embodiment of the Invention] FIG. 1 schematically shows the overall configuration of a selective smoothing filter mechanism that is an embodiment of the present invention. In Figure 1,
The smoothing filter mechanism according to the present invention includes an input device 1 that derives a pixel string constituting an image to be processed, and a smoothing processing section 3 that performs a predetermined smoothing process on pixels from the input device.

The input device 1 sequentially reads digitized pixel data to be smoothed from an imaging device such as a TV camera or a video camera, an image reading device such as a scanner, or a storage device that stores image information, and converts it into an original pixel string. Derive.

The smoothing processing unit 3 performs, for example, 8-neighborhood smoothing processing on the original pixel data (for example, luminance or density) from the input device 1. As shown in FIG. 2, this 8-neighborhood smoothing process is a smoothing process performed using the luminance data of the target pixel and the luminance data of eight pixels adjacent to the target pixel. The method of this smoothing process, that is, the weights (coefficients) of the smoothing filter are appropriately set depending on the content of the process.

The smoothing filter mechanism further includes an edge extraction unit 2 that receives the original pixel string from the input device 1 and extracts an edge region of the image.
and a neighborhood information extraction section 4 that receives pixel data from the edge extraction section 2 and extracts a region near the edge of the image.

The edge extraction unit 2, for example, P (i, j) −P (i+1, j+1)+
1 like lp (i,j+1)-P (i+1,j)
The edge region of the image is extracted using the second order differentiation, and a binarization process is performed between the extracted edge region and the remaining region. When edge extraction is performed using a first-order differential, if this differential value is greater than or equal to a certain value, it can be determined that the pixel is included in the edge region. Therefore, to classify edge areas and non-edge areas, the value obtained by this first-order differentiation is compared with a predetermined threshold, and the value of each pixel is set as "0" or "1" according to the comparison result. By showing the edge region, image information extracted from the edge region can be obtained.

In addition to first-order differentiation, the edge extraction operation performed by the edge extraction unit 2 includes, for example, a Sobel operator (an operation that indicates the correlation between the upper and lower rows of the target pixel using adjacent pixels in 3 rows and 3 columns, and It is also possible to use other edge extraction techniques.

At this time, an appropriate method is used to binarize the edge region according to the edge extraction method used.

The neighborhood information extraction unit 4 receives the pixel string from the edge extraction unit 2, and detects for each pixel whether or not there is a pixel in the edge area within a predetermined neighborhood area (for example, 8 neighborhoods).

The neighborhood information extraction unit 4 further extracts a predetermined number (for example, one) of pixel data indicating "1" in eight neighborhoods of the pixel P, for example.
If there are more than a predetermined number, data of "1" is output, and if the number is less than a predetermined number, data of "O" is output. As a result, the neighborhood information extraction unit 4 outputs an image in which edges and neighborhood regions have been extracted.

The filter mechanism further includes an output selection device 5 and an output device 6. The output selection device 5 includes the neighborhood information extraction section 4
Using the pixel data from as a control signal, either the original pixel provided from the human power device 1 or the processed pixel data provided from the smoothing processing section 3 is passed through and transmitted to the output device 6. This output selection device 5 is composed of a 2:1 multiplexer, and indicates that the pixel from the neighborhood information extraction unit 4 is an edge or neighborhood region pixel, for example, “1”.
If data is given, the original pixel from the input device 1 is passed through, and if not, the smoothed pixel data from the smoothing processing unit 3 is output.

The output device 6 is configured with a memory device, a display device, or the like, and stores or displays processing results. According to the above configuration, the neighborhood information extraction unit 4 can classify pixels into pixels that are affected by the edge area in smoothing processing and pixels that are not affected, and according to this classification, original pixels or pixels after smoothing processing can be performed. data can be output, and a smoothed image preserved with edge regions can be derived. Next, the operation of the smoothing filter shown in FIG. 1 will be explained with reference to FIG. 3, which shows the flow of the operation.

First, a pixel string representing an original image having a luminance distribution as shown in FIG. The edge extraction section 2 and the smoothing processing section 3 perform edge extraction operations and smoothing processing, respectively, after being provided with pixel data necessary for their respective processing. For example, to derive a pixel matrix with 3 rows and 3 columns, simply use two line memories (having a delay time corresponding to one row of an image) and four delay elements each having a delay time of one sampling period. It can be configured by using Such a circuit configuration is, for example, commonly used in the field of image processing.

This pixel data from the human-powered device 1 is given to an edge extraction section 2 and a smoothing processing section 3. To simplify the explanation, it is assumed that the smoothing processing section 3 performs smoothing processing by taking the average value of the pixels on the right side, and the edge extraction section 2 performs the smoothing process by taking the average value of the pixels on the right side. It is assumed that edge 7z detection is performed based on the absolute value of the difference with the luminance value. As a result, the smoothing processing unit 3 derives a pixel array as shown in FIG. 3(e).

In the extraction unit 2, a pixel array as shown in FIG. 3(b) is first obtained by an extraction operation.

This edge extraction unit 2 performs a binarization process on the edge-enhanced pixel data using a predetermined threshold value, and sets the data value of pixels in the edge area to "1" and the data value of other pixels to "0". As a result, image information containing only edge information is derived by the edge extraction section 2 and transmitted to the neighborhood information extraction section 4.

The neighborhood information extraction section 4 determines whether each pixel of the edge information image from the edge extraction section 2 includes a predetermined number of pixels included in the edge region in a predetermined neighborhood region thereof. If a pixel constituting an edge area is included in the neighboring area, that pixel is regarded as an edge neighboring area pixel and "1° pixel data is output, and for other pixels, " Outputs O° pixel data. As a result, the neighborhood information extraction unit 4 derives an image including information on the edge region and the edge vicinity region, as shown in FIG. 3(d). This pixel data from the neighborhood information extraction section 4 is given to the output selection device 5 as a control signal. The output selection device 5 selectively passes either the original pixel string from the input device 1 or the smoothed pixels from the smoothing processing section 3 according to the pixel data from the neighborhood information extraction section 4 . Therefore, from this output selection device 5,
Original pixels from the input device are output for pixels corresponding to edges and edge-near regions, and smoothed pixels from the smoothing processing section 3 are output for other pixels. As a result, a final image as shown in FIG. 3(f) is obtained.

As is clear from a comparison between FIG. 3(e) and FIG. 3(f), the smoothed image according to the present invention (the smoothed image) Figure 3 (f)) is
The edge information is preserved, and a smooth image can be obtained without losing the sharpness of the image in the edge region.

In the above operation description, the edge extraction section 2, the smoothing processing section 3, and the neighborhood information extraction section 4 each write or read the processing results into the memory for the work area, and sequentially perform the processing operations. But that's fine. That is, the edge extraction section 2 performs arithmetic processing on the original pixel information from the input device 1 after being given a predetermined number of pixels, writes the processing result to the working area memory, and similarly performs arithmetic processing on the original pixel information from the input device 1.
may read out information from the memory written by the edge extraction section 2, perform a neighborhood extraction process, and provide it to the output selection device 5 as control information. At this time, the smoothing processing result in the smoothing processing section 3 is also written to the memory, and from this memory,
The pixel data derived by the neighborhood information extraction section 4, the original pixel information given from the input device 1, and the pixel data from the smoothing processing section 3 are time-aligned and read out so that they respectively correspond to the same pixel. This allows accurate smoothing processing. The control path system in this case is not shown, or it can be easily constructed by using a buffer circuit and controlling the write and successive output timings.

Furthermore, in place of the above-mentioned configuration using a working memory, a bathophore circuit is used to buffer the information from the edge extraction unit 2, and after predetermined data (for example, pixel data for three rows) has been stored, neighboring The information extractor 4 performs predetermined processing on the data from the buffer memory, and also uses a delay circuit, for example, to adjust the output timing of the original image from the input device 1 and the processed data from the smoothing processor 3. If a buffer means is provided, the buffer circuit from the edge extraction section 2 to the neighborhood information extraction section 4 can also be configured simply using a delay circuit, so that arithmetic processing can be performed without writing to or reading from memory. This makes it possible to reduce the work memory in order to provide the working area necessary for this smoothing process and to pipeline the arithmetic processing, making it possible to perform high-speed processing.

Furthermore, in the above embodiment, as an example of smoothing processing and edge extraction, only the difference from the right side or the average value is used in FIG. 3, and only the edge region in the vertical direction of the drawing is extracted. , this is an exemplary process to simplify the explanation; edge region extraction in the horizontal direction and edge extraction in the diagonal direction can be extracted using a normal edge detection method, and the boundary region of this image can be easily extracted. Needless to say, it can be extracted accurately.

Furthermore, in the above embodiment, in order to extract the vicinity of an edge, neighboring area extraction is performed using 8 neighboring pixels, but this process is performed using, for example, 4 neighboring pixels as shown in FIG. It may be configured as follows. Here, the four neighbors refer to four pixels located on the left, right, top, and bottom of the target pixel.

Furthermore, in the above embodiment, when the input device 1 is a memory, pixel data corresponding to each pixel on the screen is stored in the memory, so that it is possible to simply specify an address without using a delay line or the like. It is also possible to easily read 8 neighboring pixel data in 3 rows and 3 columns or 4 neighboring pixel data and provide them to the smoothing processing section 3 and the edge extraction section 2.

[Effects of the Invention] As described above, according to the present invention, either the original image or the smoothed image can be selectively output based on the edge suppression area included in the original image. Since the structure is configured as follows, changes in the edge areas of the image are preserved, and the desired smoothing process is applied to most of the image that does not include the edge areas, thereby obtaining a smoothed image with clear boundaries. It becomes possible.

In addition, in the above processing, edge extraction and neighborhood information extraction perform neighborhood processing, so by having a configuration that sequentially transmits data of this desired neighborhood, it is possible to simply provide a buffer means such as a delay circuit and information extraction section,
Pipeline processing becomes possible by synchronizing only the sending timings of the original pixels from the input device and the smoothed image from the smoothing processing section, making it possible to perform image processing at high speed.

[Brief explanation of drawings]

FIG. 1 is a diagram schematically showing the overall configuration of a selective smoothing filter mechanism that is an embodiment of the present invention. FIG. 2 is a diagram illustrating 8 neighborhoods in 8 neighborhoods processing performed in smoothing processing, neighborhood information extraction, etc. FIG. 3 is a diagram showing the flow of operation of a selective smoothing filter mechanism that is an embodiment of the present invention. Figure 4 is an example of another processing method used in smoothing processing, edge extraction, etc.
FIG. 3 is a diagram illustrating neighboring pixels. FIG. 5 is a diagram showing the flow of image processing in conventional smoothing processing. In the figure, 1 is an input device, 2 is an edge extraction section, 3 is a smoothing processing section, 4 is a neighborhood information extraction section, 5 is an output selection device,
6 is an output device. In addition, in the figures, the same reference numerals indicate the same or corresponding parts.

Claims (3)

[Claims] (1) A mechanism for performing a smoothing process in which the density value of each pixel constituting a screen is expressed using the density value of the pixel and pixels included in a predetermined neighborhood area of the pixel, wherein the image on the screen is means for deriving the pixel string before the smoothing process from the information; means for applying the smoothing process to each pixel from the deriving means; means for detecting an edge region that defines an area; in response to the output of the detection means, whether smoothing processing in the smoothing processing means is performed using pixels included in the edge region detected by the detection means; means for determining for each pixel, and in response to the output of the determining means,
A selective smoothing filter mechanism based on edge information, comprising means for selectively passing a density value of a pixel from either one of the deriving means and the smoothing processing means as smoothed pixel data. (2) The selective passing means is configured to pass a signal corresponding to the certain pixel when the determining means determines that the smoothing process has been performed on the certain pixel using the pixels included in the edge area. selective smoothing based on edge information according to claim 1, wherein the pixel data from the deriving means that corresponds to the certain pixel from the smoothing processing means is passed; filter mechanism. (3) The detecting means includes means for creating an edge extracted image by subjecting the pixel string from the deriving means to edge enhancement processing that emphasizes the boundary area of the image, and the determining means includes: the edge extracted image creating means For each of the output image data from including means for determining whether or not the conversion process has been performed;
A selective smoothing filter mechanism using edge information according to claim 1.