JP2009188514A - Image encoding apparatus, image decoding apparatus and image processing program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide an image encoding apparatus capable of achieving highly efficient encoding, and an image decoding apparatus and an image processing program. <P>SOLUTION: A separating section 11 alternately separates a baseband signal per one pixel and generates separated images A and B. An orthogonal transform section 13-1 uses a DCT type 2 as orthogonal transform for the separated image A. An orthogonal transform section 13-2 uses a DST type 2 to execute orthogonal transform of a difference signal between a prediction signal of a 1/2 pixel accuracy inverse orthogonal transform section (1/2 pixel accuracy inverse DCT type 2) obtained from the separated image A and a signal of the separated image B. A variable-length encoding section 15 encodes a signal obtained by applying orthogonal transform and quantization to the image signal of the separated image A and a signal obtained by applying orthogonal transform and quantization to the difference signal of the separated image B. <P>COPYRIGHT: (C)2009,JPO&INPIT

Description

  The present invention relates to an image encoding technique in a video processing device, and more particularly to an image encoding device, an image decoding device, and an image processing program for efficiently storing a video signal in a storage device.

  Conventionally, various reversible or irreversible formats for recording digital video signals are used in VTRs, computers, digital cameras, and the like. Typical examples of non-compressed formats include BMP (bitmap format), and JPEG and MPEG as a lossy compression format. These encoding schemes realize the encoding of moving images by skillfully using human visual characteristics and actively reducing the amount of information. Basically, the most recent N × N An orthogonal transform such as DCT (Discrete Cosine Transform) or DCT (Discrete Sine Transform) is applied to a small region composed of pixels, and the orthogonal transform coefficient is quantized to obtain a variable length code. By performing the conversion, the encoding of the moving image is realized.

  In general, when the screen is divided into small areas, the difference in signal activity (complexity) between the small areas increases due to the high resolution of the video signal. This is because, by increasing the resolution of the video signal, it becomes possible to express a precise object that could not be expressed by a conventional video signal. However, at the same time, since the resolution of a flat area (an area that can be easily encoded) in the screen is increased, the signal activity in the flat area is further reduced compared to other areas. Since this flat area has a small signal property difference between adjacent areas, highly efficient compression coding can be realized by coding a plurality of areas together. For example, H.M. In standards such as the H.264 encoding method, encoding is realized by devising a small area dividing method.

  In addition, an image encoding device that separates images into partial images that do not share pixels and uses prediction by oversampling of orthogonal transforms is disclosed (for example, see Patent Document 1 and Patent Document 2).

JP 2007-0774616 A JP 2007-151062 A

  In a typical conventional coding method, a video signal is divided into small regions composed of adjacent pixels, and coding is performed for each small region, and the degree of quantization is varied depending on the characteristics of the pixels in the small region. Yes. Therefore, the file format of the BMP format is a format that does not compress the video signal, and the recording area consumption of the recording medium increases although it is high quality. On the other hand, JPEG for still images and MPEG encoding for moving images are encoding methods that exhibit high compression performance while maintaining high quality. However, if excessive compression is performed, the encoding areas are arranged in a rectangle. Therefore, the difference in the compression rate of the minimum unit of encoding tends to appear as block distortion. Although the block distortion can be reduced by increasing the minimum encoding unit, in this case, it is difficult to adaptively change the compression rate in accordance with the properties of the image, which tends to reduce the image quality. Therefore, if the methods disclosed in Patent Document 1 and Patent Document 2 are used, encoding with less block distortion is realized by using DCT and DST together.

  In particular, the method disclosed in Patent Document 2 separates an image in an encoder by pixel thinning, preferably into two, performs DCT orthogonal transform on the first separated image, and after quantization, has a variable length. In addition to encoding, the quantization coefficient is inversely quantized, and a prediction signal for predicting a signal of another separated image is generated by 1/2 pixel precision inverse DCT transform, and the prediction signal for the other separated image and the original signal are generated. DST orthogonal transform is performed on the difference, the coefficient is quantized, variable-length decoded, and each variable-length encoded signal is transmitted (reverse processing is performed on the decoding side). By using the separated image in this way, it is possible to reduce block distortion caused by the combination of DCT, DST, and quantization, and to generate a visually good encoded image.

  However, although DCT or DST ideally shows the best effect when applied to the same pixel signal, it is further necessary to reduce block distortion caused by the combination of DCT, DST, and quantization. There was room for improvement.

  Accordingly, the present invention has been made to solve such a problem, and an object of the present invention is to provide an image encoding device, an image decoding device, and an image processing program that realize high-efficiency encoding while reducing block distortion. It is to provide.

  First, in the pixel signal prediction according to the present invention, similarly to the methods disclosed in Patent Document 1 and Patent Document 2, a video signal to be encoded is separated into, for example, two separated images, and adjacent pixels with high correlation are obtained. A second separated image (interpolated signal) is generated using the first separated image (original signal) consisting of (hereinafter referred to as oversampling). This is based on the fact that the interpolation signal based on the low-frequency component of the first separated image has a high correlation with the original signal of the second separated image and is often suitable as a predicted value. In this case, the first separated image is a signal obtained by thinning out the second separated image from the original video signal, and the amount of encoded information is small. Further, since the second separated image is predicted using the first separated image composed of adjacent pixels having high correlation, the prediction is made relatively accurately, and the prediction signal of the second separated image based on the first separated image The energy of the difference signal between the second separated image and the second separated image becomes smaller. This makes it possible to realize highly efficient encoding.

  However, in the image coding system according to the present invention, in order to suppress high-frequency components due to aliasing that occur during oversampling, the correlation between the orthogonal transform base and the aforementioned interpolation signal is used only on the decoding side. . That is, only the orthogonal transform coefficient of the interpolation signal is weighted on the decoding side according to the strength of the correlation.

  Further, in the image coding method according to the present invention, the methods disclosed in Patent Document 1 and Patent Document 2 are further improved, and on the image coding side, an orthogonal transform coefficient for the first separated image (original signal). Therefore, a prediction signal of another separated image is generated without weighting, and DCT and DST type 2 orthogonal transformations are performed on each separated image. On the other hand, the image decoding side multiplies the DST type 2 coefficient by a weighting coefficient in consideration of inverse transformation and aliasing for a separated image using the DST type 2 when performing a process reverse to the image encoding process. Then, DST conversion with 1/2 pixel accuracy is performed to generate a decoded signal selected with a predetermined threshold. As a result, a bias is generated in the spectrum distribution of the orthogonal transform coefficient of the difference signal representing the error of the predicted image. Therefore, block distortion can be further reduced, and the original signal is correctly reproduced by inverse transformation, and noise due to high-frequency components is further suppressed in the signal value of the interpolation signal.

  That is, an image encoding device according to the present invention is an image encoding device that performs orthogonal transform, quantization, and variable length encoding on a pixel signal for each small region that constitutes an image of a video signal. A separation unit that separates adjacent pixels from the pixel signal and generates a plurality of separated images, and a first type of orthogonal transformation and quantization is performed on the pixel signal of the first separated image of the plurality of separated images. An inverse transform / quantization unit, and an inverse quantum that performs inverse quantization on the signal subjected to the orthogonal transform and quantization in order to predict a pixel signal of another separated image among the plurality of separated images. And a first inverse orthogonal transform that performs inverse orthogonal transform with 1 / n pixel accuracy on the orthogonal transform coefficient obtained from the inverse quantization unit as a natural number n excluding zero and predicts the other separated image And predicted by the first inverse orthogonal transform A subtractor for obtaining a difference signal between the prediction signal of the separated image and the signal of the other separated image generated by the separation unit, and performing a second type of orthogonal transformation and quantization on the difference signal A differential signal orthogonal transform / quantization unit; a signal obtained by performing orthogonal transform and quantization on the differential signal by the differential signal orthogonal transform / quantization unit; and the first signal by the orthogonal transform / quantization unit. A variable-length coding unit that performs variable-length coding on a signal that has undergone orthogonal transformation and quantization on a separated image, and when the first type is DCT type 2, the second type Is DST type 2.

  Note that the differential signal orthogonal transform / quantization unit obtains a small region orthogonal transform coefficient so that the error energy of the differential signal is equal to or less than a predetermined threshold.

  Furthermore, the image decoding apparatus of the present invention is an image decoding apparatus that decodes a signal encoded by the image encoding apparatus of the present invention, and is a variable decoding unit that performs variable length decoding on the signal encoded by the image encoding apparatus. Inverse orthogonal transform is performed on the signal of the first orthogonal transform coefficient generated by performing inverse quantization on the quantized signal of the first separated image obtained from the long decoding unit and the image encoding device. And a second inverse orthogonal transform unit that generates a decoded signal of the first separated image, and an inverse orthogonal transform is performed on the first orthogonal transform coefficient with a 1 / n pixel accuracy as a natural number n excluding zero. A third inverse orthogonal transform unit that generates a prediction signal corresponding to the other separated image, and inverse quantization on the quantized signal of the difference signal of the other separated image obtained from the image encoding device. To the signal of the second orthogonal transform coefficient generated by applying A fourth inverse orthogonal transform unit that performs the second type of inverse orthogonal transform to generate a prediction difference signal, and the second orthogonal transform coefficient includes the other separated image and the first separated image. An orthogonal transform coefficient weighting unit that takes into account the first weighting coefficient according to the correlation between the orthogonal transform coefficient weighting unit, and the result of the orthogonal transform coefficient weighting unit is subjected to inverse orthogonal transform with 1 / n pixel accuracy as the natural number n, A fifth inverse orthogonal transform unit that generates a quantization error prediction signal for predicting a quantization error of the first separated image; and the first separation included in the signal obtained from the image coding device. Using the quantization information of the image, the restriction unit restricts the quantization error prediction signal to the decoded signal of the first separated image generated by the second inverse orthogonal transform unit and the restriction unit. The quantized error prediction signal generated is added, A first adder for generating a decoded signal of the first separated image, and a fourth inverse orthogonal transform unit for the prediction signal of the other separated image generated by the third inverse orthogonal transform unit. A second adder that adds the prediction difference signal generated by the above-described step and generates a decoded signal of the other separated image, and a decoding of the new first separated image generated by the first adder. And a synthesis unit that synthesizes a signal and a decoded signal of the other separated image generated by the second addition unit.

  Furthermore, the present invention provides a computer functioning as an image encoding apparatus, separating means for separating adjacent pixels from the pixel signal for each small region and generating a plurality of separated images, a first of the plurality of separated images. Orthogonal transform / quantization means for performing a first type of orthogonal transform and quantization on the pixel signals of the separated images, the orthogonal signal for predicting pixel signals of other separated images of the plurality of separated images Inverse quantization means for performing inverse quantization on the transformed and quantized signal, and orthogonal transform coefficient obtained from the inverse quantization means is inverse orthogonal with 1 / n pixel accuracy as a natural number n excluding zero. A first inverse orthogonal transform unit that performs transformation and predicts the other separated image, a prediction signal of another separated image predicted by the first inverse orthogonal transform, and the other generated by the separating unit With separated image signal A difference signal orthogonal transform / quantization means for subjecting the difference signal to a second type of orthogonal transform and quantization, and the difference signal orthogonal transform / quantization means for the difference signal. A variable-length code that performs variable-length coding on a signal that has undergone orthogonal transformation and quantization, and a signal that has undergone orthogonal transformation and quantization on the first separated image by the orthogonal transformation / quantization means An image processing program for functioning as an image processing unit, characterized in that when the first type is DCT type 2, the second type is DST type 2 It is done.

  Furthermore, the present invention provides a computer functioning as an image decoding device for decoding a signal encoded by the image encoding device of the present invention, and a variable length decoding for variable length decoding of the signal encoded by the image encoding device. A first orthogonal transform coefficient signal generated by performing inverse quantization on the quantized signal of the first separated image obtained from the image encoding device, and performing inverse orthogonal transform on the signal A second inverse orthogonal transform unit for generating a decoded signal of the first separated image; the first orthogonal transform coefficient is subjected to inverse orthogonal transform with 1 / n pixel accuracy as a natural number n excluding zero; Third inverse orthogonal transform means for generating a prediction signal corresponding to the separated image, generated by performing inverse quantization on the quantized signal of the difference signal of the other separated image obtained from the image coding device Second orthogonal transform coefficient signal In contrast, a fourth inverse orthogonal transform unit that performs the second type of inverse orthogonal transform to generate a prediction difference signal, the second orthogonal transform coefficient, the other separated image and the first separated image An orthogonal transform coefficient weighting means that takes into account a first weighting coefficient corresponding to the correlation between the orthogonal transform coefficient weighting means, the result of the orthogonal transform coefficient weighting means is subjected to inverse orthogonal transform with 1 / n pixel accuracy as the natural number n, and Fifth inverse orthogonal transform means for generating a quantized error prediction signal for predicting a quantized error of the first separated image; and the first separated image included in the signal obtained from the image coding apparatus. Using the quantization information, the limiting means for limiting the quantization error prediction signal and the decoded signal of the first separated image generated by the second inverse orthogonal transform means are limited by the limiting means. The quantization error prediction signal Are added to the prediction signal of the other separated image generated by the first adding means for generating a new decoded signal of the first separated image and the third inverse orthogonal transform means, and the fourth inverse A second addition unit that adds the prediction difference signal generated by the orthogonal transform unit and generates a decoded signal of the other separated image, and the new first separated image generated by the first addition unit. And an image processing program for functioning as a synthesizing unit that synthesizes the decoded signal of the other separated image generated by the second adding unit.

  According to the present invention, in intra coding of a video signal, it is possible to realize coding with a higher compression rate and less deterioration due to coding, increasing the use efficiency of a recording medium and raising the utilization efficiency of a video transmission band. be able to.

  Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings. First, an image encoding device will be described.

  Usually, there are four types of DCT and DST used as orthogonal transforms depending on the discretization method (each DST type is referred to as type 1, type 2, type 3, and type 4). These types (hereinafter also referred to as orthogonal transform types) derive DFT (Discrete Fourier Transform) by the weighted sum of orthogonal transform coefficients in the same orthogonal transform type. Therefore, in the prior art method (that is, the method of Patent Document 1 and Patent Document 2 described above) and the method according to the present invention, pixels that are not actually the same are regarded as the same position pixels, and each separated image is represented by DCT and DST. By encoding and adding, it can be said that a result that is almost equivalent to the case of encoding using DFT is derived with fewer expressions than when using DFT.

  In one embodiment of the present invention, the same orthogonal transform type is used as the orthogonal transform used in each separated image, and Type 2 is used. Further, the first separated image is subjected to orthogonal transformation by DCT type 2 and the difference signal between the prediction signal of the second separated image generated using the coefficient information and the original signal of the second separated image is DST. Orthogonal transformation by type 2 is used. Further, on the decoding side, the DST type 2 orthogonal transform coefficient is multiplied by a weight. By configuring in this way, the coding efficiency especially for high frequency components is greatly improved.

  That is, the encoding transmission apparatus according to the embodiment described below predicts a signal at the pixel position of the other separated image by coding the first separated image with DCT type 2 and performing inverse DCT with 1/2 pixel accuracy. A signal is generated, and a difference signal composed of a difference between the prediction signal and a pixel signal at a corresponding pixel position of another separated image is encoded by DST type 2. Thereby, highly efficient encoding is realizable.

(Image coding device)
FIG. 1 is a block diagram showing a configuration of an encoding apparatus according to an embodiment of the present invention. The encoding apparatus 10 shown in FIG. 1 is an example in which encoding is performed by performing orthogonal transformation on an 8 × 8 pixel region of two separated images. The encoding device 10 includes a separation unit 11, an orthogonal transform unit (DCT type 2) 13-1, a quantization unit 14-1, a variable length coding unit 15, a buffer 16, and an orthogonal transform unit (DST type). 2) 13-2, quantization unit 14-2, inverse quantization unit 17, 1/2 pixel accuracy inverse orthogonal transform unit (1/2 pixel accuracy inverse DCT type 2) 18, and subtraction unit 19 I have.

  When the encoding device 10 stores a video baseband signal in a predetermined area of a frame memory (not shown), the separation unit 11 reads a baseband signal of a small area for each predetermined pixel area from the frame memory. Then, the square pixel signal of the baseband signal is alternately separated for each pixel, two separated images A and B are generated, and stored in a predetermined area of the frame memory.

  Specifically, the separation unit 11 reads 45 pixel signals in a squarely arranged (2 × N−1) × (2 × N) region (N = 8) shown in the upper part of FIG. A rotation process of ° is performed and read out as a pixel signal shown in the lower part of FIG. Then, the separation unit 11 converts the pixel signal in the frame shown in the lower part of FIG. 2 into a square pixel signal of 8 × 8 pixels for the separated image A and 8 × 8 for the separated image B as shown in FIG. The pixel is separated into square pixel signals. Thus, by performing the 45 ° rotation process, the processing load or the processing speed of the orthogonal transformation process can be improved. Instead of the 45 ° rotation process, when the separated images A and B are extracted from the baseband signal written in the frame memory, the pixel signal is read from the corresponding memory address of the frame memory so as to rotate in the 45 ° direction. Thus, it can be written in a frame memory (not shown).

  In general, as shown as separated images A and B, image encoding is performed in units of rectangular areas. When quantization is performed on this rectangular area at a different compression rate, image degradation may be significantly detected at the boundary of the area. Further, it is known that the human visual characteristics are most inferior in the sensitivity of the oblique pattern inclined by 45 ° compared to the horizontal and vertical patterns. Therefore, when encoding is performed for each rhombus region of an oblique pattern inclined by 45 °, even if the image at the boundary of the oblique pattern is deteriorated, it is difficult to be recognized in terms of human visual characteristics. That is, adopting the rhombus orthogonal transformation coding method using the separation unit 11 including 45 ° rotation processing makes the coding area a rhombus rhombus instead of a rectangle, and the boundary between adjacent areas is an oblique pattern. As a result, the amount of coding distortion detected can be reduced.

  In this way, the baseband signal of the video of, for example, a 15 × 16 pixel region is separated into an 8 × 8 pixel separated image A indicated by a black circle and an 8 × 8 pixel separated image B indicated by a white circle. To be separated. In this case, the coordinates of each pixel do not change. Separated images A and B represent minimum encoding units of partial images obtained by separating a high-resolution image (baseband signal) into two. Then, as will be described later, from the orthogonal transform coefficient of the separated image A obtained by the inverse quantization unit 17 via the orthogonal transform unit (DCT type 2) 13-1 and the quantization unit 14-1, 1/2 pixels are obtained. The accuracy inverse orthogonal transform unit (1/2 pixel accuracy inverse DCT type 2) 18 generates a prediction signal for predicting the signal at the pixel position of the separated image B. Finally, the signal for decoding the separated image B is obtained by converting the difference (difference signal) between the original signal of the separated image B and the prediction signal thereof to an orthogonal transform unit (DST type 2) 13-2 and a quantization unit 14-. 2 is generated by performing orthogonal transform and encoding via 2. Hereinafter, the components after the separation unit 11 will be specifically described.

  The orthogonal transform unit (DCT type 2) 13-1 receives the pixel signal of the 8 × 8 pixel square pixel block from the separation unit 11, and performs DCT type 2 on the 8 × 8 pixel region to perform orthogonal transform. To output orthogonal transform coefficients (DCT coefficients).

  The quantization unit 14-1 receives the buffer amount from the buffer 16, performs quantization according to the target bit rate on the DCT coefficient that is the result of DCT by the orthogonal transform unit (DCT type 2) 13-1, Output quantized coefficients. The quantization coefficient output by the quantization unit 14-1 is input to the variable length encoding unit 15 and the inverse quantization unit 17.

  The inverse quantization unit 17 receives the quantization coefficient from the quantization unit 14-1, performs inverse quantization, and outputs a DCT coefficient.

  A 1/2 pixel accuracy inverse orthogonal transform unit (1/2 pixel accuracy inverse DCT type 2) 18 receives a DCT coefficient from the inverse quantization unit 17 and is inverse orthogonal with a 1/2 pixel accuracy to the input DCT coefficient. Conversion is performed and pixel information of 16 × 16 pixels is calculated. Then, the 1/2 pixel accuracy inverse orthogonal transform unit (1/2 pixel accuracy inverse DCT type 2) 18 outputs the 16 × 16 pixel image information as a prediction signal. Here, the image information of 16 × 16 pixels output by the 1/2 pixel accuracy inverse orthogonal transform unit (1/2 pixel accuracy inverse DCT type 2) 18 is separated by the separation unit 11 from the image signal. A pixel portion corresponding to the pixel position of the generated separated image B becomes a prediction signal of the separated image B. Note that the 1/2 pixel accuracy inverse orthogonal transform unit (1/2 pixel accuracy inverse DCT type 2) 18 performs inverse orthogonal transform only for the pixel portion corresponding to the separated image B in order to reduce the amount of computation of inverse orthogonal transform. You may make it give. In addition, the 1/2 pixel accuracy inverse orthogonal transform unit (1/2 pixel accuracy inverse DCT type 2) 18 is a small region so that the error energy of the difference signal is equal to or less than a predetermined threshold value in order to maximize the compression rate. It is preferable to obtain the orthogonal transformation coefficient. In this embodiment, the 1/2 pixel accuracy inverse orthogonal transform unit (1/2 pixel accuracy inverse DCT type 2) 18 is described, but the present invention is not limited to 1/2 pixel accuracy.

  Here, DCT type 2, inverse DCT type 2, 1/2 pixel accuracy inverse DCT type 2, DST type 2, inverse DST type 2, and 1/2 pixel accuracy inverse DST type 2 are shown below. F (x, y) is a pixel signal, x and y are pixel coordinates, f ′ (x ′, y ′) is a pixel signal with ½ pixel accuracy, and x ′ and y ′ are ½ pixel accuracy. , F (u, v) is a DCT coefficient or DST coefficient (orthogonal transform coefficient) obtained by DCT or DST, u is a horizontal frequency, and v is a vertical frequency.

  The DCT type 2 conversion equation is as shown in equation (1).

  The inverse DCT type 2 conversion formula is as shown in formula (2).

  Also, the conversion formula of 1/2 pixel accuracy inverse DCT type 2 is as shown in formula (3).

  The DST type 2 conversion formula is as shown in formula (4).

  The inverse DST type 2 conversion equation is as shown in equation (5).

  Also, the conversion formula of 1/2 pixel precision inverse DST type 2 is as shown in formula (6).

  The subtracting unit 19 outputs an 8 × 8 pixel separated image B signal from the separating unit 11 and a 1/2 pixel accuracy inverse orthogonal transform unit (1/2 pixel accuracy inverse DCT type 2) 18 to a 16 × 16 pixel prediction signal. Are respectively input, and a difference signal between the signal of the 8 × 8 pixel separated image B and the prediction signal in the pixel portion corresponding to the signal is generated.

  The orthogonal transformation unit (DST type 2) 13-2 receives the difference signal generated by the subtraction unit 19 and performs orthogonal transformation on the 8 × 8 pixel region.

  In FIG. 1, the quantization unit 14-2 receives the buffer amount from the buffer 16, and sets the target bit for the orthogonal transform coefficient of the DST coefficient that is the result of the orthogonal transform by the orthogonal transform unit (DST type 2) 13-2. Quantizes according to the rate and outputs the quantized coefficients.

  The variable length coding unit 15 receives the quantized coefficients from the quantizing units 14-1 and 14-2, respectively, and for each quantized coefficient, for example, a quantized orthogonal transform coefficient defined in ISO / IEC13818-2 Is encoded by an encoding method using run-length encoding, and is output as a bit stream signal converted into a variable-length code via the buffer 16. This bit stream signal is added after the code for identifying the information of the separated image A, the variable length code of the quantization coefficient, and the variable length code of the last quantization coefficient to indicate the end of the quantization coefficient. Is added after the signal of the separated image A configured by the end code of the code, the code for identifying the information of the separated image B, the variable length code of the quantization coefficient, and the variable length code of the last quantization coefficient. A signal of the separated image B configured by an end code for indicating the end of the quantization coefficient, and information for identifying the type of orthogonal transform of DCT or DST are included.

  That is, the variable length coding unit 15 outputs a code for identifying the separated image A by variable length coding, outputs a variable length code of the quantization coefficient, and indicates the end of the quantization coefficient. Is output, a code for identifying the separated image B is output, a code for identifying orthogonal transformation is output, a variable-length code of the quantization coefficient is output, and the quantization coefficient is output. An end code for indicating the end of is output.

  As described above, according to the image encoding device 10 according to the embodiment of the present invention, the separation unit 11 generates the separated images A and B by alternately separating the baseband signal for each pixel, and has a 1/2 pixel accuracy. The inverse orthogonal transform unit (1/2 pixel accuracy inverse DCT type 2) 18 predicts the pixel signal of the separated image B using the separated image A, and the subtractor 19 predicts the pixel signal of the separated image B and the separated image B. The difference signal is output from the signal, and the orthogonal transform unit (DST type 2) 13-2 and the quantization unit 14-2 perform orthogonal transform and quantization on the difference signal by DST type 2, and variable length coding is performed. The unit 15 encodes the pixel signal of the separated image A and the difference signal of the separated image B.

  Since the prediction signal for predicting the separated image B is generated using the separated image A including the best matching pixels for the pixels of the separated image B, it can be a relatively accurate signal. As a result, the difference signal of the separated image B becomes a flat signal with low activity, and the encoding efficiency is higher than when the separated image B is encoded as it is. Further, since the pixel signal of the separated image A is a signal obtained by thinning the separated image B from the baseband signal, the amount of encoded information is halved. That is, according to the image encoding device 10, it is possible to realize high-efficiency encoding as compared with the case where the separated images A and B are encoded as they are, and further effectively reduce unnecessary high-frequency components. Can be made.

  Further, block distortion is detected by combining orthogonal transformation of the basic image signal and the prediction difference signal, in particular, orthogonal transformation of the same property as DCT type 2 and DST type 2, and a combination with high matching accuracy. Make it harder. Further, a high-frequency component that should form a high-resolution image is included in the low-resolution separated image A or B by an orthogonal transform coefficient weighting unit 46-2 of the image decoding device 40, which will be described later, that is compatible with the image encoding device 10. In spite of this, it is possible to more effectively remove high-frequency components in the interpolated signal that may be caused by oversampling, and as a result, predictive distortion (that is, high-frequency distortion) can be reduced.

  That is, the method of performing orthogonal transform on one separated image with DCT type 2 and the other separated image with DST type 2 is a process of weighting only interpolation signals to be orthogonally transformed with DST type 2 on the decoding side described later. The alignment with is excellent. This can be confirmed by actually using simulation. That is, by encoding the separated image A by DCT (orthogonal transformation using a cos component) and predicting the separated image B from the encoded signal, the difference signal of the separated image B becomes a signal mainly composed of the sin component. . Therefore, by performing DST (orthogonal transformation using sin components) on the separated image B, the power concentration degree can be improved as compared with the case where DCT is used for the orthogonal transformation of the prediction difference signal of the separated image B. Therefore, it is possible to improve the image quality of the video to be processed (particularly, the effect of reducing coding distortion such as block distortion).

  The separation unit 11 provided in the image encoding device 10 separates the baseband signal into the separated images A and B. However, for example, when separating into a rectangular shape, the separation unit 11 separates the baseband signal into four separated images. The number of images to be separated is not limited. In addition, although the separation unit 11 reads the baseband signal of a small area for each 15 × 16 pixel area from the frame memory, the reading unit is not limited to this.

  Further, the quantization unit 14-2 may not output the quantization coefficient when all the input orthogonal transform coefficients are quantized to 0. In this case, since the variable length coding unit 15 does not input the quantization coefficient from the quantization unit 14-2, it is not necessary to perform coding on the quantization unit 14-2 side. As a result, since the output bit stream signal is only the signal on the quantization unit 14-1 side, the encoded information can be substantially halved.

  Next, an image decoding apparatus according to an embodiment of the present invention will be described.

(Image decoding device)
FIG. 4 is a block diagram showing a configuration example of the decoding device according to the embodiment of the present invention. The image decoding device 40 is configured to decode the bit stream from the image encoding device 10 of the above-described embodiment, and includes a variable length decoding unit 41, inverse quantization units 42-1 and 42-2, and orthogonal Transform coefficient weighting unit 46-2, 1/2 pixel accuracy inverse orthogonal transform unit (1/2 pixel accuracy inverse DCT type 2) 43-1 and 1/2 pixel accuracy inverse orthogonal transform unit (1/2 pixel accuracy inverse DST) Type 2) 43-2, inverse orthogonal transform unit (inverse DCT type 2) 45-1, inverse orthogonal transform unit (inverse DST type 2) 45-2, adders 48 and 49, limiting unit 50, and composition Part 51.

  The variable length decoding unit 41 receives the bit stream signal from the image encoding device 10 shown in FIG. 1 and performs variable length decoding. Specifically, information for identifying the separated image A and the separated image B, quantization information, information for identifying the type of orthogonal transformation of DCT or DST, orthogonal transformation coefficient information, and the like are decoded. After the variable length decoding, the variable length decoding unit 41 sends the identified separated image A and separated image B quantized signals to the inverse quantization units 42-1 and 42-2, respectively.

  The inverse quantization unit 42-1 receives the quantized signal of the identified separated image A and performs inverse quantization to generate an orthogonal transform coefficient signal. The inverse orthogonal transform unit (inverse DCT type 2) 45- 1 and 1/2 pixel accuracy inverse orthogonal transform unit (1/2 pixel accuracy inverse DCT type 2) 43-1.

  That is, the signal input to the 1/2 pixel accuracy inverse orthogonal transform unit (1/2 pixel accuracy inverse DCT type 2) 43-1 is used for decoding the separated image B, and the inverse orthogonal transform unit (inverse DCT) is used. Type 2) The signal input to 45-1 is used for decoding the separated image A.

  The inverse orthogonal transform unit (inverse DCT type 2) 45-1 receives the signal of the orthogonal transform coefficient of the separated image A from the inverse quantization unit 42-1, and performs inverse orthogonal transform (for example, inverse DCT represented by Expression (2)). To generate a decoded signal of the separated image A and send it to the adder 48.

  The 1/2 pixel accuracy inverse orthogonal transform unit (1/2 pixel accuracy inverse DCT type 2) 43-1 receives the signal of the orthogonal transform coefficient of the separated image A from the inverse quantization unit 42-1, and outputs 1/2 pixel. Inverse orthogonal transformation with accuracy is performed (for example, inverse DCT type 2 with 1/2 pixel accuracy shown in Expression (3)). As a result, a prediction signal corresponding to the pixel position of the separated image B is generated as image information of 16 × 16 pixels and sent to the adding unit 49. Here, when the variable length decoding unit 41 starts decoding from the beginning of the bit stream signal, the variable length decoding unit 41 decodes the quantization information of the separated image A and the quantized orthogonal transform coefficient information. Is done. In this way, the inverse quantization unit 42-1 and the 1/2 pixel accuracy inverse orthogonal transform unit (1/2 pixel accuracy inverse DCT type 2) 43-1, for example, the 8 × 8 DCT coefficient of the separated image A. A signal of image information of 16 × 16 pixels is generated from the signal, and an 8 × 8 prediction signal corresponding to the pixel position of the separated image B can be generated.

  The inverse quantization unit 42-2 performs inverse quantization on the differential signal for decoding the identified separated image B that has been variable-length decoded by the variable-length decoding unit 41, and signals of orthogonal transform coefficients of the differential signal Is transmitted to the orthogonal transform coefficient weighting unit 46-2 and the inverse orthogonal transform unit (inverse DST type 2) 45-2.

That is, the signal input to the orthogonal transform coefficient weighting unit 46-2 is used for generating orthogonal transform coefficients (for decoding the separated image A) weighted by weighting coefficients w _{u and v} shown in Table 1 described later. The signal input to the inverse orthogonal transform unit (inverse DST type 2) 45-2 is used for generating orthogonal transform coefficients (for decoding the separated image B) that do not use the weighting coefficients w _{u, v} .

  The inverse orthogonal transform unit (inverse DST type 2) 45-2 receives the signal of the orthogonal transform coefficient of the difference signal of the separated image B from the inverse quantization unit 42-2, and performs inverse orthogonal transform (for example, shown in Expression (5)). A differential signal (hereinafter referred to as a prediction differential signal) for decoding the separated image B by performing inverse DST type 2) is generated and sent to the adding unit 49.

The orthogonal transform coefficient weighting unit 46-2 uses a predetermined weighting coefficient (Table 1) for the signal of the orthogonal transform coefficient of the difference signal of the separated image B input from the inverse quantization unit 42-2. Then, an orthogonal transform coefficient X ′ [u] [v] corresponding to Equation (8) described later is calculated, and the 1/2 pixel accuracy inverse orthogonal transform unit (1/2 pixel accuracy inverse DST type 2) 43-2 is calculated. Send it out. Here, a matrix composed of the weighting coefficients w _{u, v} is stored in advance in a memory (such as a ROM) not shown on the image decoding apparatus side. Preferably, the image decoding device 40 uses a matrix composed of weighting coefficients w _{u, v} included in the bit stream signal sent from the image encoding device 10 in advance.

  The effect of the orthogonal transform coefficient weighting unit 46-2 will be described in detail. In principle, in order to completely represent the pixels of the separated image A and the separated image B, a 16 × 16 orthogonal transform coefficient is required, and the other signal (an isolated image in the embodiment) with an 8 × 8 orthogonal transform coefficient. In order to express the signal of the pixel position of B), there is insufficient information on the high frequency component. For this reason, when a decoded signal of the separated image A is simply generated from the prediction signal corresponding to the separated image B sent from the image encoding device 10, high-frequency distortion (that is, aliasing distortion) of the high-frequency signal component originally included in the image information is generated. This is not preferable because it is included in the prediction signal in relatively many cases.

  That is, this is due to a base shift between the pixel block of the separated image A and the pixel block corresponding to the pixel position of the separated image B obtained by oversampling the separated image A. In order to realize more efficient image compression, it is preferable that the high-frequency component is weaker as the spectral distribution of the image signal. Therefore, it is preferable to be able to remove such high-frequency distortion on the decoding side. Become. Therefore, when generating the decoded signal of the separated image A from the prediction signal corresponding to the separated image B, the orthogonal transformation coefficient weighting unit 46-2 weights the frequency component of the orthogonal transformation coefficient (DST coefficient), and 1/2 The high-frequency distortion is reduced by generating a prediction signal by the pixel accuracy inverse orthogonal transform unit (1/2 pixel accuracy inverse DST type 2) 43-2.

  Hereinafter, a method for deriving weights to orthogonal transform coefficients by the orthogonal transform coefficient weighting unit 46-2 will be described.

  First, the base shift regarding the separated image A and the separated image B will be described. FIG. 5 shows a DCT base pattern composed of black circle pixels (separated image A) in FIG. FIG. 6 shows the DCT at the position of the white circle pixel (separated image B) interpolated from the DCT base pattern composed of the black circle pixels (separated image A) of FIG. Indicates the base pattern. Both are 8 × 8 pixel DCT base patterns composed of only white and black circles. The base patterns shown in FIGS. 5 and 6 having substantially the same frequency components are patterns whose phases are shifted by 1/2 pixel in the vertical and horizontal directions, respectively. Therefore, each base pattern has the same frequency component of the signal constituting it, and both are correlated.

  However, comparing each base pattern, it can be seen that the correlation is low because the pixel shift increases with respect to the period of the frequency constituting the base as the high frequency component is obtained. Such a tendency can be grasped as a general moving image property. Similarly, when a separated image is generated with a coarser inter-pixel distance, the correlation between the base patterns tends to decrease as the spatial deviation increases. Therefore, since the image encoding device 10 of the present embodiment transmits the encoded data of each separated image with such a tendency, it is necessary for the image decoding device 40 to suitably suppress this high frequency component. become.

Since the image coding apparatus 10 according to the present embodiment uses DST type 2 as orthogonal transform of the difference signal of the separated image B, the (u, v) -th base pattern of the DST at the pixel position of the separated image A is represented by fu _, Let _V [x] [y] and the DST coefficient obtained by the DST type 2 be X [u] [v]. The inverse DST of X [u] [v] is oversampled, the base pattern of the pixel position of the obtained separated image B is predicted as f ′ _{u, v} [x] [y], and the pixel position of the separated image B is predicted. A new DST coefficient for this is X ′ [u] [v]. (U, v) The weighting coefficient w _{u, v} of the next coefficient is obtained by the following equation (7). Note that N is the number of pixels (N = 8 for an 8 × 8 pixel separated image), x and y are pixel position coordinates, u is a horizontal frequency, and v is a vertical frequency.

Also, the DST coefficient X ′ [u] [v] is obtained from the weighting coefficient w _{u, v} and the DST coefficient X [u] [v] using the equation (8).

A matrix of weighting coefficients w _{u, v} using DST type 2 as orthogonal transform of the difference signal of the separated image B is expressed as shown in Table 1.

  The matrix shown in Table 1 sets the accuracy according to the target performance, is stored in advance in a memory (for example, ROM) in the image encoding device or the image decoding device, and is used when weighting the DST coefficients.

  As described above, the orthogonal transform coefficient weighting unit 46-2 calculates the basis between the pixel block of the separated image A and the pixel block corresponding to the pixel position of the separated image B obtained by oversampling the separated image A. A pre-defined matrix of weighting factors (Table 1) for correcting the deviation is used.

  The 1/2 pixel accuracy inverse orthogonal transform unit (1/2 pixel accuracy inverse DST type 2) 43-2 inputs the signal X ′ [u] [v] of the orthogonal transform coefficient from the orthogonal transform coefficient weighting unit 46-2. Then, inverse orthogonal transform with 1/2 pixel accuracy is performed (for example, inverse DST type 2 with 1/2 pixel accuracy shown in Expression (6)). As a result, a signal for predicting the quantization error of the separated image A (hereinafter referred to as a quantization error prediction signal) is generated as image information of the difference signal of 16 × 16 pixels, and is sent to the restriction unit 50.

  Here, when the variable length decoding unit 41 starts decoding from the beginning of the bit stream signal, the variable length decoding unit 41 continues to the quantization information of the separated image A and the quantized orthogonal transform coefficient information. Then, the quantized orthogonal transform coefficient information of the difference signal of the separated image B is decoded. In this way, the inverse quantization unit 42-2 and the 1/2 pixel accuracy inverse orthogonal transform unit (1/2 pixel accuracy inverse DST type 2) 43-2 perform 8 × 8 orthogonality of the difference signal of the separated image B. An image information signal having a difference of 16 × 16 pixels is generated from the transform coefficient, and an 8 × 8 quantization error prediction signal corresponding to the pixel position of the separated image A can be generated.

  Note that the 1/2 pixel accuracy inverse orthogonal transform unit (1/2 pixel accuracy inverse DST type 2) 43-2 may perform the inverse transform only for the pixels corresponding to the separated images A and B. As a result, the number of inverse transform operations can be reduced to half.

  The limiting unit 50 receives the decoded signal of the separated image A generated by the inverse orthogonal transform unit (inverse DCT type 2) 45-1, and also outputs a 1/2 pixel accuracy inverse orthogonal transform unit (1/2 pixel accuracy inverse DST). Type 2) Input the quantization error prediction signal of the separated image A from 43-2, compare both, and if the quantization error prediction signal of the separated image A is a signal within a predetermined quantization range, the separation A quantization error prediction signal for image A is output, and if the signal is not within a predetermined quantization range, a signal based on the quantization information for separated image A is output.

  FIG. 7 is a block diagram showing a configuration of the limiting unit 50 shown in FIG. The configuration of the restriction unit 50 is the same as the orthogonal transformation unit 61 of the orthogonal transformation unit (DCT type 2) 13-1 of the image encoding device 10 used for the separated image A (ie, DCT type 2). The inverse orthogonal transform unit 63 is provided. Specifically, the restriction unit 50 includes a DCT (DCT type 2) unit 61, a comparison unit 62, and an IDCT (inverse DCT type 2) unit 63.

  The DCT (DCT type 2) unit 61 receives the quantization error prediction signal (8 × 8 pixel difference) of the separated image A from the 1/2 pixel accuracy inverse orthogonal transform unit (1/2 pixel accuracy inverse DST type 2) 43-2. Image information), is subjected to orthogonal transformation (DCT type 2), and outputs DCT coefficients.

The comparison unit 62 receives the quantization information of the separated image A from the variable length decoding unit 41 and the DCT coefficient from the DCT (DCT type 2) unit 61, and compares the two. When the quantization value for each component in the quantization information is Q [u] [v] and the value of each component in the DCT coefficient is DCT [u] [v], for example, the following comparison restriction is performed.
for (u = 0; u <8; u ++) {
for (v = 0; v <8; v ++) {
if (DCT [u] [v]> Q [u] [v] / a) DCT [u] [v] = Q [u] [v] / a;
}
}
Here, a is a variable preset by system design, for example, a = 2.0. That is, the DCT (DCT type 2) unit 61 outputs Q [u] [v] / a (limited DCT coefficient) when DCT [u] [v]> Q [u] [v] / a. When DCT [u] [v] ≦ Q [u] [v] / a, DCT [u] [v] (DCT coefficient) is output.

  The IDCT (inverse DCT type 2) unit 63 receives the DCT coefficient or the limited DCT coefficient from the comparison unit 62, performs inverse orthogonal transform (inverse DCT type 2), converts the pixel information, and outputs the pixel information to the addition unit 48. To do.

  In FIG. 4, an adder 48 receives the decoded signal of the separated image A and the pixel information signal from the restricting unit 50 from the inverse orthogonal transform unit (inverse DCT type 2) 45-1, and adds them. Then, a decoded signal of a new separated image A is output. That is, the adding unit 48 adds the quantized error prediction signal of the separated image A or the limited quantization error predicted signal to the decoded signal of the separated image A, and sends the new decoded image A decoded signal to the combining unit 51. Send it out. In addition, the adding unit 49 includes a prediction signal of the separated image B from the 1/2 pixel accuracy inverse orthogonal transform unit (1/2 pixel accuracy inverse DCT type 2) 43-1, and an inverse orthogonal transform unit (inverse DST type 2). The prediction difference signal of the separated image B is input from 45-2, and these are added, and the decoded signal of the separated image B is output. That is, the adding unit 49 adds the prediction difference signal of the separated image B to the predicted signal of the separated image B predicted from the separated image A, and outputs the decoded signal of the separated image B to the combining unit 51.

  The synthesizing unit 51 receives the decoded signal of the new separated image A from the adding unit 48 and the decoded signal of the separated image B from the adding unit 49, and synthesizes these decoded signals with the original image arrangement signal. And stored as a baseband signal in a memory (not shown).

  As described above, according to the image decoding device 40 according to the embodiment of the present invention, the decoded signal of the separated image A is generated using the quantization error prediction signal of the separated image A or the limited quantization error prediction signal. The decoded signal of the separated image B is generated using the prediction difference signal of the separated image B. Therefore, in the case of the image decoding device 40 according to the above-described embodiment, the decoded signal generated from the DST coefficient that has been weighted only on the decoding side is used for the reconstruction of the original image. Can be obtained.

  The present invention has been described with reference to the embodiments. However, the present invention is not limited to the above embodiments, and various modifications can be made without departing from the technical idea thereof. For example, the above embodiment has described the encoding method when the pixel density to be encoded is halved. That is, in the image encoding device 10, the separation unit 11 generates separated images A and B having a pixel density of 1/2, and a 1/2 pixel accuracy inverse orthogonal transform unit (1/2 pixel accuracy inverse DCT type 2) 18. ½ pixel inverse orthogonal transform. On the other hand, the separation unit 11 generates a separation image having a pixel density of 1/4, 1/16, etc., and uses a variable density inverse orthogonal transform unit (1 / n pixel inverse orthogonal transform (not shown) as a natural number n excluding zero. Part) may perform pixel inverse orthogonal transform such as 1/4, 1/16, and the like. In this case, the subtraction unit 19 generates a difference signal between the prediction signal from the 1 / n pixel inverse orthogonal transform unit and the separated image signal, and the variable length encoding unit 15 repeatedly uses the difference signal. By encoding, variable density orthogonal transform encoding can be realized.

  For example, a 1 / n pixel inverse orthogonal transform unit that replaces the 1/2 pixel accuracy inverse orthogonal transform unit (1/2 pixel accuracy inverse DCT type 2) 18 shown in FIG. And y only in increments of 1 / n, this can be realized by setting x to x / n and y to y / n. In this case, the 1 / n pixel precision inverse conversion equation is as shown in Equation (9).

  Also, the inverse DST of 1 / n pixels can be derived similarly.

  In the above-described embodiment, the two separated images are shown as the closest separation method for separating the separated images. For example, the separated images can be divided into four separated images having a pixel arrangement as shown in FIG. Also, in this case, the weighting coefficient derivation method may be determined by interpolating the signal at the pixel position of another separated image by oversampling of the orthogonal transformation base and correlation between the signal and the orthogonal transformation base, as described above. .

  Furthermore, as one aspect of the present invention, the image encoding device 10 or the image decoding device 40 can be configured as a computer that functions as each device. A program for causing a computer to realize each of the above-described components (such as an orthogonal transform unit) is stored in a storage unit provided inside or outside each computer. Such a storage unit can be realized by an external storage device such as an external hard disk or an internal storage device such as ROM or RAM. The control unit provided in each computer can be realized by control of a central processing unit (CPU) or the like. In other words, the CPU can appropriately read from the storage unit a program in which the processing content for realizing the function of each component is described, and realize the function of each component on the computer. Here, the function of each component may be realized by all or part of the hardware.

  In the embodiment described above, the program describing the processing contents for realizing the function of each component can be recorded on a computer-readable recording medium. As the computer-readable recording medium, any recording medium such as a magnetic recording device, an optical disk, a magneto-optical recording device, and a semiconductor memory may be used.

  In addition, the program describing the processing contents can be distributed by selling, transferring, or lending a portable recording medium such as a DVD or CD-ROM, and such a program can be distributed on a network such as an IP. The program can be distributed by storing the program in the storage unit of the server and transferring the program from the server to another computer via the network.

  In addition, a computer that executes such a program can temporarily store, for example, a program recorded on a portable recording medium or a program transferred from a server in its own storage unit. As another embodiment of the program, the computer may directly read the program from a portable recording medium and execute processing according to the program, and each time the program is transferred from the server to the computer. In addition, the processing according to the received program may be executed sequentially. Note that the program in this aspect includes information provided for processing of an electronic computer and equivalent to the program (data that is not a direct command to the computer but has a property that defines the processing of the computer). .

  Although the above embodiments have been described as representative examples, it will be apparent to those skilled in the art that many changes and substitutions can be made within the spirit and scope of the invention. Accordingly, the invention should not be construed as limited by the embodiments described above, but only by the claims.

  According to the present invention, it is possible to realize high-compression encoding in intra-encoding of a video signal, so that it is possible to increase the usage efficiency of the recording medium and increase the usage efficiency of the video transmission band. This is useful for recording media and / or recording systems that handle moving images, and broadcast systems.

It is a block diagram which shows the structure of the image coding apparatus of one Example by this invention. It is a figure explaining 45 degree rotation of the pixel signal for rhombus orthogonal transformation. It is a figure explaining isolation | separation of the pixel signal for a rhombus orthogonal transformation. It is a block diagram which shows the structure of the image decoding apparatus of one Example by this invention. It is a figure which shows an example of the DCT base pattern of the separated image A. It is a figure which shows an example of the DCT base pattern corresponding to the separated image B interpolated by the oversampling by 1/2 pixel precision inverse DCT of the separated image A. It is a block diagram which shows the structure of a restriction | limiting part. It is a figure which shows the pixel arrangement | positioning divided | segmented into four separated images.

Explanation of symbols

DESCRIPTION OF SYMBOLS 10 Image coding apparatus 11 Separation part 13-1 Orthogonal transformation part (DCT type 2)
13-2 Orthogonal transformation unit (DST type 2)
14-1 Quantization unit 14-2 Quantization unit 15 Variable length encoding unit 16 Buffer 17 Inverse quantization unit 18 1/2 pixel accuracy inverse orthogonal transform unit (1/2 pixel accuracy inverse DCT type 2)
19 Subtraction unit 40 Image decoding device 41 Variable length decoding unit 42-1, 42-2 Inverse quantization unit 43-1 1/2 pixel accuracy inverse orthogonal transform unit (1/2 pixel accuracy inverse DCT type 2)
43-2 1/2 Pixel Accuracy Inverse Orthogonal Transformer (1/2 Pixel Accuracy Inverse DST Type 2)
45-1 Inverse Orthogonal Transformer (Inverse DCT Type 2)
45-2 Inverse Orthogonal Transformer (Inverse DST Type 2)
46-2 Orthogonal transform coefficient weighting unit 48, 49 Adder unit 50 Limit unit 51 Combining unit 61 DCT unit 62 Comparison unit 63 IDCT unit

Claims (4)

An image encoding device that performs orthogonal transform, quantization, and variable length encoding on a pixel signal for each small region constituting an image of a video signal,
A separation unit that separates adjacent pixels with respect to the pixel signal for each small region and generates a plurality of separated images;
An orthogonal transform / quantization unit that performs a first type of orthogonal transform and quantization on a pixel signal of a first separated image of the plurality of separated images;
An inverse quantization unit that performs inverse quantization on the signal subjected to the orthogonal transformation and quantization in order to predict a pixel signal of another separated image of the plurality of separated images;
A first inverse orthogonal transform unit that performs inverse orthogonal transform with 1 / n pixel accuracy on the orthogonal transform coefficient obtained from the inverse quantization unit as a natural number n excluding zero, and predicts the other separated image;
A subtractor for obtaining a difference signal between a prediction signal of another separated image predicted by the first inverse orthogonal transform and a signal of the other separated image generated by the separation unit;
A differential signal orthogonal transform / quantization unit for performing a second type of orthogonal transform and quantization on the differential signal;
The difference signal orthogonal transform / quantization unit performs orthogonal transform and quantization on the difference signal, and the orthogonal transform / quantization unit performs orthogonal transform and quantization on the first separated image. A variable-length coding unit that performs variable-length coding on the applied signal,
When the first type is DCT type 2, the second type is DST type 2, and the image coding apparatus is characterized in that: An image decoding apparatus for decoding a signal encoded by the image encoding apparatus according to claim 1,
A variable-length decoding unit for variable-length decoding the signal encoded by the image encoding device;
The first orthogonal transform coefficient signal generated by performing inverse quantization on the quantized signal of the first separated image obtained from the image encoding device is subjected to inverse orthogonal transform and the first orthogonal transform coefficient is generated. A second inverse orthogonal transform unit that generates a decoded signal of the separated image of
A third inverse orthogonal transform unit configured to perform inverse orthogonal transform with 1 / n pixel accuracy on the first orthogonal transform coefficient as a natural number n excluding zero and generate a prediction signal corresponding to the other separated image;
The inverse of the second type is applied to the signal of the second orthogonal transform coefficient generated by performing inverse quantization on the quantized signal of the difference signal of the other separated image obtained from the image encoding device. A fourth inverse orthogonal transform unit that performs orthogonal transform to generate a prediction difference signal;
An orthogonal transform coefficient weighting unit that adds a first weighting coefficient corresponding to the correlation between the other separated image and the first separated image to the second orthogonal transform coefficient;
The result of the orthogonal transform coefficient weighting unit is subjected to inverse orthogonal transform with 1 / n pixel accuracy as the natural number n to generate a quantization error prediction signal for predicting the quantization error of the first separated image. A fifth inverse orthogonal transform unit;
A limiting unit that limits the quantization error prediction signal using quantization information of the first separated image included in a signal obtained from the image encoding device;
The quantized error prediction signal limited by the limiting unit is added to the decoded signal of the first separated image generated by the second inverse orthogonal transform unit, and a new decoded signal of the first separated image is added. A first adder for generating
The prediction difference signal generated by the fourth inverse orthogonal transform unit is added to the prediction signal of the other separated image generated by the third inverse orthogonal transform unit, and the decoded signal of the other separated image is added. A second adder for generating
A combining unit that combines the decoded signal of the new first separated image generated by the first adding unit and the decoded signal of the other separated image generated by the second adding unit;
An image decoding apparatus comprising: A computer functioning as an image encoding device,
Separating means for separating adjacent pixels for the pixel signal for each small region and generating a plurality of separated images;
Orthogonal transform / quantization means for performing a first type of orthogonal transform and quantization on a pixel signal of a first separated image of the plurality of separated images;
Inverse quantization means for performing inverse quantization on the signals subjected to the orthogonal transformation and quantization in order to predict pixel signals of other separated images among the plurality of separated images;
A first inverse orthogonal transform unit that performs an inverse orthogonal transform with 1 / n pixel accuracy as a natural number n excluding zero with respect to the orthogonal transform coefficient obtained from the inverse quantization unit, and predicts the other separated image;
A subtractor for obtaining a difference signal between a prediction signal of another separated image predicted by the first inverse orthogonal transform and a signal of the other separated image generated by the separating unit;
A differential signal orthogonal transform / quantization means for subjecting the differential signal to a second type of orthogonal transform and quantization; and
The difference signal orthogonal transform / quantization means performs orthogonal transform and quantization on the difference signal, and the orthogonal transform / quantization means performs orthogonal transform and quantization on the first separated image. An image processing program for causing a given signal to function as variable length coding means for performing variable length coding,
An image processing program characterized in that when the first type is DCT type 2, the second type is DST type 2. A computer functioning as an image decoding device that decodes a signal encoded by the image encoding device according to claim 1,
Variable length decoding means for variable length decoding the signal encoded by the image encoding device;
The first orthogonal transform coefficient signal generated by performing inverse quantization on the quantized signal of the first separated image obtained from the image encoding device is subjected to inverse orthogonal transform and the first orthogonal transform coefficient is generated. Second inverse orthogonal transform means for generating a decoded signal of the separated image of
A third inverse orthogonal transform unit that performs inverse orthogonal transform with 1 / n pixel accuracy on the first orthogonal transform coefficient as a natural number n excluding zero and generates a prediction signal corresponding to the other separated image;
The inverse of the second type is applied to the signal of the second orthogonal transform coefficient generated by performing inverse quantization on the quantized signal of the difference signal of the other separated image obtained from the image encoding device. A fourth inverse orthogonal transform unit that performs orthogonal transform to generate a prediction difference signal;
Orthogonal transform coefficient weighting means for adding a first weighting coefficient corresponding to the correlation between the other separated image and the first separated image to the second orthogonal transform coefficient;
The result of the orthogonal transform coefficient weighting unit is subjected to inverse orthogonal transform with 1 / n pixel accuracy as the natural number n to generate a quantization error prediction signal for predicting the quantization error of the first separated image. Fifth inverse orthogonal transform means;
Limiting means for limiting the quantization error prediction signal using quantization information of the first separated image included in the signal obtained from the image encoding device;
The quantized error prediction signal limited by the limiting unit is added to the decoded signal of the first separated image generated by the second inverse orthogonal transform unit, and a new decoded signal of the first separated image is added. First adding means for generating
The prediction difference signal generated by the fourth inverse orthogonal transform unit is added to the prediction signal of the other separated image generated by the third inverse orthogonal transform unit, and the decoded signal of the other separated image Second adding means for generating
Function as a combining unit that combines the decoded signal of the new first separated image generated by the first adding unit and the decoded signal of the other separated image generated by the second adding unit. Image processing program.

Images