JPH09187000A - Image processor, image processing method, image coder, image coding method and recording medium - Google Patents

Abstract

PROBLEM TO BE SOLVED: To prevent coding rate from being deteriorated by detecting pixels forming a periodic area based on an orthogonal transformation coefficient obtained by applying orthogonal transformation to small areas including each pixel. SOLUTION: Image data to be coded are fed to a 2-dimension characteristic point detection circuit 2, a quantization device 4, a computing element 10, and a periodic area detection circuit 20 via a preliminary filter 1. In the circuit 20, a DCT circuit 21 at first applies orthogonal transformation such as discrete cosine transformation DCT to a noted picture element in the line scanning order, for example, that is, a picture element included in a small area of 7×7 picture elements. Then an absolute sum calculation circuit 22 calculates an absolute sum of DCT coefficients in the small area, a square sum calculation circuit 23 calculates a square sum of DCT coefficients and they are outputted to a computing element 24. Thus, picture elements forming a periodic area are detected to prevent deterioration in the coding rate.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image processing device and an image processing method, an image coding device and an image coding method, and a recording medium. In particular, an image processing apparatus and an image processing method capable of easily detecting a periodic area, which is an area that changes periodically in an image,
The present invention relates to an image encoding device, an image encoding method, and a recording medium.

[0002]

2. Description of the Related Art Image coding methods are roughly classified into
There are an encoding method that handles an image in pixel units and an encoding method that handles an image in block units.

As a coding method for handling an image on a pixel-by-pixel basis, there is, for example, a predictive coding method.

Further, as an encoding method for handling an image in block units, an orthogonal transform encoding method such as DCT (discrete cosine transform), a band division method typified by subband encoding using wavelet transform, and further, , There is an encoding method in which the band-divided image is divided into blocks and handled (hereinafter, these are collectively referred to as a block encoding method).

The predictive coding system has a drawback that it is easy to implement it, but if a high compression rate is to be obtained, the image quality is significantly deteriorated. In addition, regarding the block coding method, by appropriately quantizing each block, even if a high compression rate is set, a relatively high-quality image can be obtained, while the compression rate is further increased. If the value is increased, there is a drawback that the visual unfavorable effects such as block distortion become remarkable.

Therefore, as a coding method in which distortion that causes visual interference does not become noticeable even at a high compression rate, at a characteristic point of the structure of an image (for example, a point forming an outline of an object). There is a coding method (structure extraction coding method by detecting structural points of an image) that extracts (detects) a certain characteristic point and efficiently codes an image based on the characteristic point.

FIG. 7 shows the configuration of an example of a conventional image coding apparatus for coding an image by such a coding method. Image data (digital data) to be encoded is input to a pre-filter 1 including a smoothing filter for removing noise. In pre-filter 1,
After the noise is removed from the image data and the necessary processing for detecting feature points is performed by the two-dimensional feature point detection circuit 102 in the subsequent stage, the two-dimensional feature point detection circuit 102, the quantizer 4, and It is output to the calculator 10.

The two-dimensional feature point detection circuit 102 detects feature points from the image data from the prefilter 1. Then, the two-dimensional feature point detection circuit 102 sequentially sets the pixels forming the image as a target pixel in, for example, the so-called line scan order, and when the target pixel is a feature point, outputs the feature point data having a value of 1, Otherwise, the feature point data having a value of 0 is output. This feature point data is supplied to the chain configuration circuit 3.

When the chain construction circuit 3 receives the feature point data for one frame from the two-dimensional feature point detection circuit 102, it performs chain coding based on the feature point data. That is, in the chain configuration circuit 3, the feature point data is 1
, A certain pixel, that is, a certain feature point as a starting point, and a feature point adjacent thereto is detected. Further, it detects a feature point adjacent to the feature point just detected, which has not been detected yet (hereinafter referred to as an undetected adjacent feature point),
Hereinafter, similar processing is performed until the undetected adjacent feature points cannot be detected. In this way, the chain configuration circuit 3 detects a series of feature points, that is, a chain, and lists the coordinates of the feature points that form the chain (hereinafter, appropriately referred to as a feature point list. ) Is generated.

The chain construction circuit 3 constructs a feature point list for all chains existing in one frame of image data, and stores it in the amplitude value determination circuit 5, the interpolator 7,
And to the multiplexer (MUX) 12.

On the other hand, the quantizer 4 quantizes the image data from the pre-filter 1, and the resulting quantized value is output to the amplitude value determination circuit 5. Amplitude value determination circuit 5
Selects only the quantized values supplied from the quantizer 4 for the characteristic points described in the characteristic point list from the chain configuration circuit 3 (hereinafter, appropriately referred to as characteristic point quantized value). Output to the inverse quantizer 6 and the multiplexer 12.

The dequantizer 6 dequantizes the feature point quantized value, and the pixel value of the feature point obtained as a result is output to the interpolator 7. The interpolator 7 reconstructs the original image based on the feature points and the pixel values at the feature points.
That is, in the interpolator 7, the corresponding image data (pixel value) from the inverse quantizer 6 is arranged at the position of the feature point based on the feature point list from the chain forming circuit 3. Further, in the interpolator 7, the pixel values of points (pixels) other than the feature points are interpolated by using the pixel values of the feature points close to the point, for example, by a method such as diffusion filtering or linear interpolation. An image obtained by the interpolation by the interpolator 7 (an image generated by the interpolation) (hereinafter, appropriately referred to as a reconstructed image) is output to the calculator 10.

Here, this reconstructed image is shown in FIG.
This is the same as that obtained by the interpolation processing of the interpolator 34 in the image decoding apparatus of.

The arithmetic unit 10 calculates the difference between the original image data and the reconstructed image from the interpolator 7 to generate a difference image. This difference image is supplied to the difference encoding circuit 50.

The feature point list output from the chain forming circuit 3 is also supplied to the area dividing circuit 8 in addition to the amplitude value determining circuit 5, the interpolator 7, and the multiplexer 12. The area dividing circuit 8 divides the image to be encoded into predetermined areas based on the feature point list from the chain forming circuit 3. That is, the area dividing circuit 8 divides the image into areas surrounded by the feature points described in the feature point list from the chain forming circuit 3 according to, for example, a snake algorithm. Further, the area dividing circuit 8 uses a unique number for each texture area (this number is set to the address of the frame memory 9 corresponding to the pixels forming the area surrounded by the feature points (hereinafter, appropriately referred to as a texture area)). Since it is for identifying the texture area,
Hereinafter, the area number will be stored as appropriate).

In the area dividing circuit 8, the image frame is treated as a feature point if necessary.

The differential encoding circuit 50 encodes the differential image from the arithmetic unit 10. That is, the difference image from the arithmetic unit 10 is input to the conversion circuit 51 and the quantization step determination circuit 52. In the transform circuit 51, the difference image is subjected to, for example, wavelet transform to be subjected to subband coding, and the transform coefficient obtained as a result is output to the quantizer 53.

On the other hand, the quantization step determination circuit 52 recognizes the texture area by referring to the frame memory 52, and supplies the variance of the pixel value of each texture area (pixel value of the difference image) from the arithmetic unit 10. Calculation is performed using the difference image. Then, the quantization step size for quantizing the transform coefficient corresponding to each texture region is determined based on the variance.

Here, the difference image is obtained by subtracting the reconstructed image from the original image data, that is, from the original image,
For example, since the contour portion (edge portion) of the object and the portion in the vicinity thereof are removed, the image is relatively uniform for each texture region. Therefore, the variance of each texture region does not have such a large value.

The quantizer 53 quantizes the transform coefficient supplied from the transform circuit 51 in units of each texture area according to the quantization step size supplied from the quantization step determination circuit 52. The VLC circuit (variable length coding circuit) 54 performs the quantization result on the variable length coding that combines, for example, Huffman coding and zero run length coding, and a multiplexer (MUX) 55.
Is output to The multiplexer 55 includes the VLC circuit 5
4, the variable length coding result (hereinafter, appropriately referred to as variable length coded data) is supplied, and the quantization step size determined by the quantization step determining circuit 52 is also supplied. In the multiplexer 55, the output of the VLC circuit 54 and the quantization step size are multiplexed and output to the multiplexer 12.

The multiplexer 12 multiplexes the feature point list from the chain forming circuit 3, the feature point quantized value from the amplitude value determining circuit 5, and the output of the multiplexer 55, and the resulting multiplexed data is stored in a buffer memory. It outputs to 13. In the buffer memory 13, the multiplexed data from the multiplexer 12 is temporarily stored, the data amount is smoothed by this, and then transmitted through the transmission path or recorded in a recording medium (not shown).

Next, FIG. 8 shows the configuration of an example of an image decoding apparatus for decoding the image coded by the image coding apparatus of FIG. The data encoded by the image encoding device in FIG. 7 is temporarily stored in the buffer memory 31, and the data amount is smoothed by this, and then the demultiplexer (D
MUX) 32. The demultiplexer 32 is
The feature point list and the feature point quantized value are separated from the input data, and the feature point list is output to the interpolator 34 and the region dividing circuit 35, and the feature point quantized value is output to the inverse quantizer 33, respectively. . Further, the demultiplexer 32 is
The remaining data is output to the differential decoding circuit 60.

In the inverse quantizer 33, the inverse quantizer 6 of FIG.
In the same manner as in the above, the quantized value of the feature point is dequantized, and the pixel value at the resulting feature point is interpolated by the interpolator 3
4 is output. In the interpolator 34, the demultiplexer 3
Based on the feature point list from 2 and the pixel value of the feature point from the inverse quantizer 33, the reconstructed image is generated by performing interpolation in the same manner as in the interpolator 7 of FIG. Output to the container 38.

At the same time, the area dividing circuit 35 divides the image into areas (texture areas) surrounded by the characteristic points based on the characteristic point list, as in the case of the area dividing circuit 8 in FIG. Further, the area dividing circuit 35
The area number is stored in the frame memory 36 configured similarly to the frame memory 9 of FIG. 7 in the same manner as in the area dividing circuit 8.

On the other hand, the difference decoding circuit 60 decodes the difference image. That is, the output of the demultiplexer 32 is supplied to the demultiplexer (DMUX) 61, where it is separated into variable length coded data and a quantization step size. The variable length coded data is IVLC.
The quantization step size is output to the circuit (variable length decoding circuit) 62 and to the inverse quantizer 63.

In the IVLC circuit 62, the variable length coded data is variable length decoded. The quantized transform coefficient obtained as a result is supplied to the inverse quantizer 63. The inverse quantizer 63 refers to the frame memory 36 to recognize the texture area, and further, to determine the quantization step size for inverse quantizing the transform coefficient of each texture area,
The quantization step size supplied from the demultiplexer 61 is selected. Then, the transform coefficient of each texture region is inversely quantized according to the selected quantization step size, and the transform coefficient obtained as a result is inversely transformed by the inverse transform circuit 6.
4 is output. The inverse transform circuit 64 subjects the transform coefficient from the inverse quantizer 63 to a transform process reverse to that of the transform circuit 51 of FIG. 7, that is, here, performs an inverse wavelet transform, and thereby a difference image is obtained. Decrypt. This difference image is output to the calculator 38.

The arithmetic unit 38 adds the reconstructed image from the interpolator 34 and the difference image from the difference decoding circuit 60 (inverse conversion circuit 64), thereby decoding and outputting the original image. It

[0028]

By the way, by the structure extraction coding method as described above, for example, a periodically changing region such as a striped pattern (hereinafter, appropriately referred to as a periodic region)
When encoding an image that includes, the information about the feature points (feature point list and the feature point list as well as the contour of the object and the periodically changing part of the image, such as a striped pattern, are detected as the feature points. There is a problem that the point quantization value) increases and the coding efficiency decreases.

The present invention has been made in view of such a situation, and makes it possible to easily detect a periodic region from an image, and further to detect a feature point from such a periodic region. By not doing so, it is possible to prevent a decrease in image coding efficiency.

[0030]

According to another aspect of the present invention, there is provided an image processing apparatus, comprising: an orthogonal transforming means for orthogonally transforming a small area including pixels constituting an image and outputting an orthogonal transform coefficient; and an orthogonal transforming means. And a detection unit that detects pixels forming a periodic region based on the output orthogonal transformation coefficient.

According to the image processing method of the fourth aspect, a small area including each pixel forming an image is orthogonally transformed, and a pixel forming a periodic area is detected based on an orthogonal transformation coefficient obtained as a result. It is characterized by

According to a fifth aspect of the present invention, an image encoding apparatus orthogonally transforms a small area including each pixel forming an image and outputs an orthogonal transform coefficient, and an orthogonal transform output from the orthogonal transform means. Pixel detection means for detecting a pixel forming a periodic area that is a periodically changing area in the image based on the coefficient, and a periodic area detected by the pixel detection means among the pixels forming the image And a reconstructing means for producing a reconstructed image on the basis of the feature points detected by the feature point detecting means. The reconstructed image generated by the means and the difference between the image are calculated,
It is characterized by comprising a difference image generating means for generating a difference image and an encoding means for encoding the difference image generated by the difference image generating means.

In the image coding method according to the sixth aspect, a small area including each pixel forming an image is orthogonally transformed, and the orthogonal transformation coefficient in the image changes periodically based on the obtained orthogonal transformation coefficient. Detecting pixels forming a periodic area, which is an area, among the pixels forming the image, excluding the pixels forming the periodic area, detecting feature points, and based on the feature points, It is characterized in that a reconstructed image is generated, a difference image between the reconstructed image and the image is generated, and the difference image is encoded.

According to a seventh aspect of the present invention, in the coded data, the coded data orthogonally transforms a small area including each pixel forming the image, and based on the resulting orthogonal transformation coefficient, the periodicity in the image is obtained. Pixels that form a periodic area that is a region that changes to are detected, and among the pixels that form the image, those that exclude the pixels that form the periodic area are detected, and feature points are detected. A reconstructed image is generated based on the reconstructed image, a difference image between the reconstructed image and the image is generated, and the difference image is encoded to obtain the image.

In the image processing apparatus according to the first aspect, the orthogonal transformation means orthogonally transforms a small area including each pixel forming an image and outputs an orthogonal transformation coefficient, and the detection means:
Based on the orthogonal transform coefficient output from the orthogonal transform means,
Pixels forming a periodic area are detected.

In the image processing method according to the fourth aspect, a small area including each pixel forming an image is orthogonally transformed,
Pixels forming the periodic region are detected based on the resulting orthogonal transform coefficient.

In the image coding apparatus according to the fifth aspect, the orthogonal transformation means orthogonally transforms a small area including each pixel forming an image, outputs an orthogonal transformation coefficient, and the pixel detection means, the orthogonal transformation means. Pixels forming a periodic area, which is an area that changes periodically in the image, are detected based on the orthogonal transform coefficient output from the means. The feature point detecting means detects the feature points of the pixels forming the image excluding the pixels forming the periodic area detected by the pixel detecting means, and the reconstructing means detects the feature points. A reconstructed image is generated based on the feature points detected by the means. The difference image generating means calculates a difference between the reconstructed image generated by the reconstructing means and the image to generate a difference image, and the encoding means encodes the difference image generated by the difference image generating means. It is designed to do.

In the image coding method according to the sixth aspect, a small area including each pixel forming an image is orthogonally transformed,
Based on the resulting orthogonal transform coefficient, pixels that form a periodic region that is a periodically changing region in the image are detected, and among the pixels that form the image, the pixels that form the periodic region are detected. The feature points are detected for the removed ones, a reconstructed image is generated based on the feature points, a difference image between the reconstructed image and the image is generated, and the difference image is encoded. Has been done.

According to a seventh aspect of the present invention, in a recording medium, a small area including each pixel forming an image is orthogonally transformed, and a periodically changing area in the image is based on an orthogonal transformation coefficient obtained as a result. Of the pixels that form the periodic area are detected, and among the pixels that form the image excluding the pixels that form the periodic area, the feature points are detected, and based on the feature points, Generate a composition image, and reconstruct the image,
Encoded data obtained by generating a difference image with the image and encoding the difference image is recorded.

[0040]

DESCRIPTION OF THE PREFERRED EMBODIMENTS Embodiments of the present invention will be described below, but before that, in order to clarify the correspondence between each means of the invention described in the claims and the following embodiments. The features of the present invention are described as follows by adding a corresponding embodiment (however, an example) in parentheses after each means.

That is, the image processing apparatus according to claim 1 is
An image processing apparatus for detecting pixels forming a periodic area which is a periodically changing area in an image, and orthogonally transforms a small area including each pixel constituting an image, and outputs an orthogonal transformation coefficient. Orthogonal transformation means (for example, the DCT shown in FIG. 2)
(Discrete cosine transform circuit 21 and the like) and detection means (for example, absolute value sum calculation circuit 22 shown in FIG. 2) that detects pixels forming a periodic region based on the orthogonal transformation coefficient output from the orthogonal transformation means. , Sum of squares calculation circuit 23, calculator 2
4 and a threshold processing circuit 25, etc.).

In the image processing apparatus according to the third aspect, the detecting means calculates the absolute value sum of the orthogonal transform coefficients in the small area (for example, the absolute value sum calculating circuit 22 shown in FIG. 2 or the like). ), A sum of squares calculation means for calculating the sum of squares of orthogonal transformation coefficients in a small area (for example, a sum of squares calculation circuit 23 shown in FIG. 2), and a sum of squares output from the sum of squares calculation means. And a division means (for example, the arithmetic unit 24 shown in FIG. 2) for dividing by the sum of absolute values output from the sum of absolute values calculation means, and the periodic region is configured based on the division result of the division means. It is characterized by detecting a pixel to be changed.

An image coding apparatus according to claim 5 is an image coding apparatus for coding a difference image between an image and a reconstructed image reconstructed based on the feature points of the image,
An orthogonal transformation unit (for example, the DCT circuit 21 shown in FIG. 2) that orthogonally transforms a small area including each pixel forming an image and outputs an orthogonal transformation coefficient, and an orthogonal transformation coefficient output from the orthogonal transformation unit. Then, a pixel detection unit (for example, the absolute value sum calculation circuits 22 and 2 shown in FIG. 2) that detects pixels forming a periodic area that is a periodically changing area in the image.
Multiply-sum calculation circuit 23, calculator 24, and threshold processing circuit 2
5) and the pixels forming the image excluding the pixels forming the periodic area detected by the pixel detecting means, and detecting the characteristic points (for example, in FIG. 1). A two-dimensional feature point detection circuit 2 shown in FIG. 1), a reconstructing unit that generates a reconstructed image based on the feature points detected by the feature point detecting unit (such as the interpolator 7 shown in FIG. 1), A difference image generating unit that calculates a difference between the reconstructed image generated by the configuring unit and the image and generates a difference image (for example, the arithmetic unit 10 shown in FIG. 1);
Encoding means for encoding the difference image generated by the difference image generating means (for example, the difference encoding circuit 50 shown in FIG. 1).
Etc.) and are provided.

Of course, this description does not mean that each means is limited to those described above.

FIG. 1 shows the configuration of an embodiment of an image coding apparatus to which the present invention is applied. In the figure, the same reference numerals are given to portions corresponding to the case in FIG. That is, this image coding apparatus is provided with a two-dimensional feature point detection circuit 2 in place of the two-dimensional feature point detection circuit 102, and a periodic area detection circuit 20 is newly provided. 7 has the same configuration as the image coding apparatus.

The two-dimensional feature point detection circuit 2 detects the feature points of the image in the same manner as the two-dimensional feature point detection circuit 102 of FIG. However, the two-dimensional feature point detection circuit 2 is based on a periodic area pixel signal output from a periodic area detection circuit 20 described later, and among the pixels that form the image from the pre-filter 1, pixels that form a periodic area. The feature points are detected for the objects excluding. Therefore, the two-dimensional feature point detection circuit 2 does not detect feature points from the periodic area.

The periodic area detection circuit 20 is, for example, as shown in FIG. 2, a DCT circuit 21, an absolute value sum calculation circuit, a square sum calculation circuit 23, a calculator 24, and a threshold value processing circuit 25.
It is composed of An image output from the pre-filter 1 is supplied to the DCT circuit 21, in which a small area including each pixel forming the image is set, and DCT processing is performed on each small area. It is designed to be applied. That is, the DCT circuit 21 centers on each pixel forming the image from the prefilter 1, for example, 7 × 7.
DCT is performed on a square small area of a pixel.

The DCT coefficient (orthogonal transform coefficient) obtained by DCTing the small area for each pixel is output to the absolute value sum calculation circuit 22 and the square sum calculation circuit 23. The absolute value calculation circuit 22 calculates the absolute value sum of the DCT coefficients of each small area and outputs it to the calculator 24. The sum of squares calculation circuit 23 calculates the sum of squares of the DCT coefficients of each small area and outputs the sum to the calculator 24.

The computing unit 24 calculates the sum of squares of the DCT coefficients from the sum of squares calculation circuit 23 as D from the sum of absolute values calculation circuit 22.
It divides by the sum of the absolute values of the CT coefficients, and supplies the divided value obtained as a result to the threshold processing circuit 25.
The threshold value processing circuit 25 compares the division value from the arithmetic unit 24 with a predetermined threshold value, and when the division value is larger than the predetermined threshold value or less than the predetermined threshold value, for example, 1 for each. Or, 0 is used as the periodic area pixel detection signal and 2
The data is output to the dimensional feature point detection circuit 2.

Next, the operation of the image coding apparatus shown in FIG. 1 will be described. The image data to be encoded is the pre-filter 1
Is supplied to the two-dimensional feature point detection circuit 2, the quantizer 4, the calculator 10, and the periodic area detection circuit 20 via the.

In the periodic area detection circuit 20, first, DC
In the T circuit 21, each pixel forming the image from the pre-filter 1 is set as a pixel of interest in, for example, a so-called line scan order, and pixels included in a small area of 7 × 7 pixels centered on the pixel of interest are selected. DCT is performed on the target.

The DCT coefficient obtained by DCTing a small area centered on the pixel of interest is the absolute value sum calculation circuit 2
2 and the sum of squares calculation circuit 23. The absolute value calculation circuit 22 calculates the sum of absolute values of the DCT coefficients in the small region, and the square sum calculation circuit 23 calculates the sum of squares of the DCT coefficients in the small region, both of which are output to the calculator 24. To be done. In the arithmetic unit 24, the DCT from the square sum calculation circuit 23
The sum of squares of the coefficients is divided by the sum of the absolute values of the DCT coefficients from the absolute value sum calculation circuit 22, and the resulting division value (=
(Sum of squares) / (sum of absolute values)) is supplied to the threshold processing circuit 25.

The threshold processing circuit 25 determines whether or not the pixel of interest constitutes a periodic area based on the division value from the calculator 24 (pixels constituting the periodic area are detected. ). That is, in the threshold processing circuit 25, the division value from the calculator 24 is compared with a predetermined threshold. When the divided value is larger than the predetermined threshold value, the threshold processing circuit 25 outputs the periodic area pixel signal whose value is 1,
When the divided value is less than or equal to the predetermined threshold value, the periodic area pixel signals having a value of 0 are output to the two-dimensional feature point detection circuit 2.

Here, the pixel of interest for which the periodic area pixel signal is 0 constitutes a periodic area (this will be described later).

In the two-dimensional feature point detection circuit 2, the pixels forming the image from the pre-filter 1 excluding the pixels for which the periodic area pixel signal is 0 (pixels forming the periodic area), That is, the feature points are detected only for the pixels having the periodic region pixel signal of 1.

Thereafter, the same processing as in the case of FIG. 7 is performed, and the resulting multiplexed data (encoded data) is obtained.
Are transmitted via the transmission path. Alternatively, it is supplied and recorded on the recording medium 14 (FIG. 1) which is, for example, an optical disc, a magneto-optical disc, a magnetic tape, an optical card or the like.

Next, it will be described that the pixel of interest for which the periodic area pixel signal is 0 constitutes a periodic area. When DCT is performed on the periodic region, the resulting D
The CT coefficient has the property of being concentrated on a specific basis.
This is because the base of the DCT is formed by the cosine waves of all frequencies in the region where the DCT is performed. Here, in FIG. 3 (A) or FIG. 3 (B) respectively,
D that is DCT of a non-periodic region or a periodic region
The CT results are shown (in the figure, the region of 4 × 4 pixels is DCT
Shows what has been processed). In the same figure (A), DCT
While the coefficients are dispersed throughout,
It can be seen that the DCT coefficients are concentrated at the position of the third column from the left on the top row.

Further, when the DCT coefficients are concentrated on a specific basis, the sum of squares thereof becomes larger than the sum of absolute values thereof. That is, the sum of absolute values or the sum of squares of the DCT results of FIG. 3A is 16, and there is no difference between them, whereas the sum of absolute values of the DCT results of FIG. The sums of squares are 4 or 16, respectively, and the sum of squares is larger than the sum of absolute values.

Therefore, for example, as described above, the DCT
By comparing the division value obtained by dividing the sum of squares of coefficients by the sum of absolute values thereof with a predetermined threshold value, it is possible to determine whether or not the pixel of interest is within the periodic region.
That is, in the embodiment of FIG. 3, the division value in the case of FIG. 3A is 1 (= 16/16), and the division value in the case of FIG. 3B is 4 (= 16/4). By setting the predetermined threshold value to, for example, 2 or 3, it is possible to detect the pixels forming the periodic area. That is, in this case, when the division value is larger than the predetermined threshold value, the pixel of interest is a pixel forming the periodic area.

In the present embodiment, the sum of squares of the DCT coefficient is divided by the sum of absolute values thereof. Therefore, when the division value is larger than a predetermined threshold value, the pixel of interest has a periodic area. The pixel is a constituent pixel. For example, when the sum of absolute values of DCT coefficients is divided by the sum of squares thereof, when the division value is smaller than a predetermined threshold value, the pixel of interest is selected. , When it is a pixel forming a periodic area.

By the conventional image coding apparatus (FIG. 7),
For example, a feature point is detected from an image in which the wall surface has a vertical striped pattern (which changes periodically) as shown in FIG.
When the area division is performed, area division as shown in FIG. 5 is obtained. In this case, in particular, it can be seen that the amount of information on the vertical striped portion of the wall surface is enormous.

On the other hand, when the image coding apparatus of FIG. 1 detects feature points from the image shown in FIG. 4 and performs area division, area division as shown in FIG. 6 is obtained. In this case, since the vertical striped portion of the wall surface is not detected as a feature point, it can be seen that the amount of information is significantly reduced as compared with the case of FIG.

Therefore, the vertical striped portion of the wall surface is coded as a differential image in the differential coding circuit 50, and the coding efficiency is degraded when the image including the periodic area is coded. Can be prevented.

The image coding apparatus to which the present invention is applied has been described above, but the image coding apparatus can be applied to both moving images and still images.

The image coding apparatus as described above is an image transmission apparatus for transmitting an image to a remote place by using a transmission medium having a limited transmission capacity, or recording an image on a recording medium, for example, a VTR. It is applicable to (video tape recorder), disk device, and others.

Further, the data coded by this image coding apparatus can be decoded by the conventional image decoding apparatus described in FIG.

Further, in the present embodiment, the DCT is performed in the periodic area detection circuit 20, but the periodic area detection circuit 20 may be configured to perform the Fourier transform and other orthogonal transformation in addition to the DCT. good.

Further, in this embodiment, the sum of squares of the DCT coefficient is divided by the sum of the absolute values thereof, and
Although the determination is made as to whether or not the pixel of interest is a pixel forming a periodic area, this determination can also be made based on, for example, the position of the base where DCT coefficients are concentrated. .

Further, in the present embodiment, DCT is performed on a small area of 7 × 7 pixels centering on the pixel of interest.
The area targeted for CT is not limited to a small area having a size of 7 × 7 pixels.

[0070]

According to the image processing apparatus of the first aspect and the image processing method of the fourth aspect, the small area including each pixel forming the image is orthogonally transformed, and the resulting orthogonal transformation coefficient is obtained. The pixels forming the periodic region are detected based on Therefore, it becomes possible to easily detect the periodic region.

According to the image coding apparatus of the fifth aspect and the image coding method of the sixth aspect, the small region including each pixel forming the image is orthogonally transformed, and the resulting orthogonal transformation coefficient is obtained. The pixels forming the periodic area, which is a periodically changing area in the image, are detected on the basis of
Then, among the pixels forming the image, the feature points are detected for those excluding the pixels forming the periodic region. Therefore, it is possible to prevent the deterioration of the coding efficiency.

According to a seventh aspect of the present invention, in a recording medium, a small area including each pixel forming an image is orthogonally transformed, and a periodically changing area in the image is obtained based on an orthogonal transformation coefficient obtained as a result. Of the pixels that form the periodic area are detected, and among the pixels that form the image excluding the pixels that form the periodic area, the feature points are detected, and based on the feature points, Generate a composition image, and reconstruct the image,
Encoded data obtained by generating a difference image with the image and encoding the difference image is recorded. Therefore, it becomes possible to record a large amount of data.

[Brief description of the drawings]

FIG. 1 is a block diagram illustrating a configuration of an embodiment of an image encoding device to which the present invention has been applied.

FIG. 2 is a block diagram showing a configuration example of a periodic area detection circuit 20 of FIG.

FIG. 3 is a DC of a non-periodic region and a periodic region.
It is a figure which shows T result.

FIG. 4 is a diagram showing an example of an image including a periodic region.

5 is a diagram showing a region division result obtained by detecting a feature point from the entire image of FIG. 4 and performing region division based on the feature point.

6 is a diagram showing a region division result obtained by detecting a feature point from a portion excluding a periodic region in the image of FIG. 4 and performing region division based on the feature point.

FIG. 7 is a block diagram showing a configuration of an example of a conventional image encoding device.

FIG. 8 is a block diagram showing a configuration of an example of a conventional image decoding device.

[Explanation of symbols]

 2 Two-dimensional feature point detection circuit 3 Chain configuration circuit 4 Quantizer 5 Amplitude value determination circuit 6 Inverse quantizer 7 Interpolator 8 Area division circuit 9 Frame memory 10 Arithmetic unit 11 Differential encoding circuit 14 Recording medium 20 Periodic area Detection circuit 21 DCT circuit 22 Absolute value sum calculation circuit 23 Square sum calculation circuit 24 Operator 25 Threshold processing circuit 50 Differential encoding circuit

Claims (7)

[Claims] 1. An image processing apparatus for detecting pixels forming a periodic area which is a periodically changing area in an image, wherein a small area including each pixel forming the image is orthogonally transformed,
An image processing, comprising: an orthogonal transformation unit that outputs an orthogonal transformation coefficient; and a detection unit that detects a pixel forming the periodic region based on the orthogonal transformation coefficient output from the orthogonal transformation unit. apparatus. 2. The image processing apparatus according to claim 1, wherein the orthogonal transformation unit transforms the small area by discrete cosine transformation. 3. The detection means calculates a sum of absolute values of the orthogonal transformation coefficients in the small area, and a sum of squares calculates a sum of squares of the orthogonal transformation coefficients in the small area. A division result of the division means for dividing the sum of squares output from the sum of squares calculation means by the sum of absolute values output from the sum of absolute value calculation means; The image processing apparatus according to claim 1, wherein the pixels forming the periodic area are detected based on the above. 4. An image processing method for detecting pixels forming a periodic area which is a periodically changing area in an image, wherein a small area including each pixel forming the image is orthogonally transformed, An image processing method, which detects a pixel forming the periodic region based on an orthogonal transformation coefficient obtained as a result. 5. An image encoding apparatus for encoding a difference image between an image and a reconstructed image reconstructed based on characteristic points of the image, the subregion including each pixel forming the image. Orthogonally transform
An orthogonal transform unit that outputs an orthogonal transform coefficient and a pixel that forms a periodic region that is a periodically changing region in the image is detected based on the orthogonal transform coefficient output from the orthogonal transform unit. Pixel detecting means, and a feature point detecting means for detecting the feature points, excluding the pixels forming the periodic area detected by the pixel detecting means among the pixels forming the image, Based on the feature points detected by the feature point detection means, a reconstruction means for generating the reconstructed image, a difference between the reconstructed image generated by the reconstructing means, and the image is calculated. An image coding apparatus comprising: a difference image generation unit that generates the difference image; and a coding unit that codes the difference image generated by the difference image generation unit. 6. An image encoding method for encoding a difference image between an image and a reconstructed image reconstructed based on feature points of the image, the subregion including each pixel constituting the image. Orthogonally transform
Based on the resulting orthogonal transform coefficient, the pixels forming a periodic area that is a periodically changing area in the image are detected, and the periodic area is formed among the pixels forming the image. The feature points are detected for the objects excluding the pixels, the reconstructed image is generated based on the feature points, a difference image between the reconstructed image and the image is generated, and the difference is calculated. An image encoding method characterized by encoding an image. 7. A recording medium on which at least coded data obtained by coding a difference image between an image and a reconstructed image reconstructed based on a feature point of the image is recorded, the coded data Is an orthogonal transformation of a small area including each pixel forming the image,
Based on the resulting orthogonal transform coefficient, the pixels forming a periodic area that is a periodically changing area in the image are detected, and the periodic area is formed among the pixels forming the image. The feature points are detected for the objects excluding the pixels, the reconstructed image is generated based on the feature points, a difference image between the reconstructed image and the image is generated, and the difference is calculated. A recording medium characterized by being obtained by encoding an image.