JP5000012B2 - Moving picture encoding apparatus, moving picture decoding apparatus, moving picture encoding method, and moving picture decoding method - Google Patents

Description

  The present invention relates to a moving image encoding apparatus and moving image encoding method for encoding a moving image with high efficiency, a moving image decoding apparatus and a moving image decoding method for decoding a moving image encoded with high efficiency, and It is about.

For example, in motion picture coding standard schemes such as “ITU-TH.264 ISO / IEC_MPEG-4AVC”, motion compensation prediction interframe coding is used as one of the coding modes.
As a prediction model of motion compensated prediction interframe coding, a model is adopted that has the highest prediction efficiency when the brightness does not change in the time direction.
However, in natural images, the brightness of the image may change from frame to frame. In particular, when the subject moves at the same time as the brightness of the image, a large amount of code is required to maintain the image quality. There is a problem.

In order to solve the above-described problem, in the moving picture encoding device disclosed in Patent Document 1 below, at least one prepared in advance when performing motion compensation prediction encoding using a motion vector From among the plurality of combinations of the reference image number and the prediction parameter, one combination is selected for each encoding target block of the input moving image signal, and the prediction image signal is determined according to the reference image number and the prediction parameter of the selected combination. Is generated.
Then, a prediction error signal representing an error of the prediction image signal with respect to the input moving image signal is generated, and the prediction error signal, motion vector information, and index information indicating the selected combination are encoded.

JP 2004-7377 A (paragraph number [0014], FIG. 1)

  Since the conventional moving image encoding apparatus is configured as described above, it is necessary to encode the same number of pieces of index information as the number of blocks to be encoded, and a large improvement in encoding efficiency cannot be expected. For this reason, when a natural image in which the subject moves at the same time as the brightness of the image is encoded, there is a problem that a large amount of code is required to maintain the image quality.

The present invention has been made to solve the above-described problems. Even when a natural image in which a subject moves is encoded at the same time as the brightness of an image changes, it is highly efficient without causing deterioration in image quality. It is an object of the present invention to obtain a moving picture coding apparatus and a moving picture coding method capable of realizing coding.
Another object of the present invention is to obtain a moving image decoding apparatus and a moving image decoding method capable of decoding a moving image encoded with high efficiency.

  In the moving picture encoding apparatus according to the present invention, the inter-frame prediction unit generates an enlarged reference image in decimal precision units by interpolating the local decoded image output from the local decoded image output unit, and the enlarged reference image For the pixel value in the reference block obtained by sub-sampling in integer precision units starting from the position indicated by the motion vector referred to in motion search, for the pixel having a specific phase with respect to the integer component of the motion vector An adaptive prediction image generation unit that adds and subtracts a predetermined offset value set in advance and outputs an image of the reference block as an adaptive prediction image, and an adaptive prediction image and an image output from the adaptive prediction image generation unit Evaluate the error of the macroblock image output from the dividing means, and output the motion vector when the evaluation is optimal as inter-frame prediction information Both, in which so as to provide a luminance change motion detecting section for outputting an adaptive prediction image when the evaluation is optimized as a prediction image.

  According to this invention, the inter-frame prediction unit generates an enlarged reference image in decimal precision units by interpolating the local decoded image output from the local decoded image output unit, and moves with respect to the enlarged reference image. The pixel value in the reference block obtained by subsampling in integer precision units starting from the position pointed to by the motion vector referenced in the search is set in advance for the pixel having a specific phase with respect to the integer component of the motion vector. An adaptive prediction image generation unit that adds and subtracts a predetermined offset value and outputs the image of the reference block as an adaptive prediction image, an adaptive prediction image output from the adaptive prediction image generation unit, and an output from the image dividing unit The error of the generated macroblock image is evaluated, and the motion vector when the evaluation is optimal is output as inter-frame prediction information. When a natural image in which the subject moves at the same time as the brightness of the image changes is encoded, because the brightness change motion detection unit that outputs an adaptive predicted image when the valence is optimal is output as a predicted image However, there is an effect that high-efficiency encoding can be realized without causing deterioration of image quality.

It is a block diagram which shows the moving image encoder by Embodiment 1 of this invention. It is a block diagram which shows the brightness | luminance change motion compensation prediction part 7 of the moving image encoder by Embodiment 1 of this invention. It is a block diagram which shows the moving image decoding apparatus by Embodiment 1 of this invention. It is a flowchart which shows the processing content of the moving image encoder by Embodiment 1 of this invention. It is a flowchart which shows the processing content of the moving image decoding apparatus by Embodiment 1 of this invention. It is explanatory drawing which shows the example of a production | generation of the expansion reference image of a quarter pixel precision. It is explanatory drawing which shows an example of the bit stream for transmitting the threshold value which determines the presence or absence of a brightness change.

Hereinafter, in order to explain the present invention in more detail, modes for carrying out the present invention will be described with reference to the accompanying drawings.
Embodiment 1 FIG.
1 is a block diagram showing a moving picture coding apparatus according to Embodiment 1 of the present invention.
In FIG. 1, when a moving image signal is input in units of frames or fields, an image dividing unit 1 divides each frame constituting a moving image into macro blocks of a predetermined size, and images of the macro blocks (hereinafter referred to as “macro blocks”). (Referred to as “block image”). The image dividing unit 1 constitutes an image dividing unit.

The subtraction unit 2 obtains a difference between the macroblock image output from the image division unit 1 and the prediction image output from the prediction image changeover switch 9 (from the pixel values of the pixels constituting the macroblock image, the pixels constituting the prediction image) The pixel value of the pixel corresponding to the pixel constituting the macroblock image is subtracted to obtain a difference), and a process of outputting a prediction error signal indicating the difference is performed.
The quantization conversion unit 3 performs a conversion process and a quantization process on the prediction error signal output from the subtraction unit 2, and performs a process of outputting a quantization conversion coefficient of the prediction error signal.

Here, as the conversion processing, for example, processing such as orthogonal frequency conversion such as DCT or FFT and orthogonal conversion such as wavelet conversion is applicable, and any conversion is possible as long as the energy distribution of the input signal is easily biased. This process is also applicable.
As the quantization processing, any quantization can be applied as long as the quantization can be expressed with a smaller set for a given signal, such as scalar quantization or vector quantization.
The subtraction unit 2 and the quantization conversion unit 3 constitute a quantization conversion unit.

The inverse quantization transform unit 4 performs a process of inversely quantizing and inverse transforming the quantized transform coefficient output from the quantization transform unit 3 to obtain a decoded prediction error signal and outputting the decoded prediction error signal.
The addition unit 5 adds the decoded prediction error signal output from the inverse quantization conversion unit 4 and the prediction image output from the prediction image change-over switch 9 and outputs a local decoded image as a result of the addition. To do. At this time, the clipping process may be performed so that the pixel value of the locally decoded image obtained as a result of the addition falls within a predetermined range.
The inverse quantization transform unit 4 and the addition unit 5 constitute a local decoded image output unit.

The frame memory 6 is a memory for storing the locally decoded image output from the adding unit 5.
The luminance change motion compensation prediction unit 7 reads out the locally decoded image stored in the frame memory 6 as an inter-frame prediction reference image, and uses the inter-frame prediction reference image and the macroblock image output from the image division unit 1, By performing inter-frame prediction, an inter-frame prediction image is generated and output, and a process of outputting inter-frame prediction information for specifying the inter-frame prediction method is performed.
That is, the luminance change motion compensation prediction unit 7 uses the inter-frame prediction reference image, and the block division information, the motion vector, and the reference frame number (a plurality of reference images are determined when the evaluation value of the difference from the macro block image is minimized). Information for specifying an image to be referred to in a certain case) is output as inter-frame prediction information, and an image region specified by the inter-frame prediction information is output as an inter-frame prediction image.
Note that the luminance change motion compensation prediction unit 7 constitutes an inter-frame prediction unit.

The intra prediction unit 8 reads out the locally decoded image stored in the frame memory 6 as an intra-frame prediction reference image, and uses the intra-frame prediction reference image to calculate the difference from the macroblock image output from the image division unit 1. The prediction mode when the evaluation value is minimized is output as intra-frame prediction information, and the process of outputting the prediction image generated by the intra prediction method specified by the intra-frame prediction information is performed as the intra-frame prediction image .
Intra prediction includes H.264. Similar to H.264, prediction in pixel units may be used, and any method can be applied to the present invention as long as the prediction is based on the peripheral information of the encoding target block. Further, even a technique for predicting a quantized transform coefficient itself such as intra prediction in MPEG-4 is applicable to the present invention.

The prediction image changeover switch 9 is an inter-frame prediction image output from the luminance change motion compensation prediction unit 7 or an intra-frame prediction output from the intra prediction unit 8 according to the macroblock prediction mode information output from the encoding control unit 10. A process for selecting one of the images and outputting the selected image as a predicted image is performed.
The encoding control unit 10 determines the optimum macroblock prediction mode and quantization parameter based on various conditions, and performs a process of outputting the optimum macroblock prediction mode and quantization parameter as macroblock mode information.

The entropy encoding unit 11 outputs the quantized transform coefficient output from the quantizing transform unit 3, the interframe prediction information output from the luminance change motion compensation prediction unit 7, and the intraframe prediction information output from the intra prediction unit 8. Then, the macro block mode information output from the encoding control unit 10 is compressed by a predetermined entropy encoding method, thereby executing processing for outputting encoded data.
Here, as the entropy coding method, any coding method can be applied to the present invention as long as it is a lossless coding method such as Huffman coding, adaptive Huffman coding, arithmetic coding, and adaptive arithmetic coding. is there.
The entropy encoding unit 11 constitutes an encoding unit.

FIG. 2 is a block diagram showing the luminance change motion compensation prediction unit 7 of the moving picture coding apparatus according to Embodiment 1 of the present invention.
In FIG. 2, the adaptive prediction image generation unit 21 reads out a reference image identified by the reference frame number included in the motion search information from among the locally decoded images stored in the frame memory 6 as an inter-frame prediction reference image. By interpolating the inter-frame prediction reference image, an enlarged reference image in decimal precision unit is generated, and a motion vector (vector to be referred to in motion search) included in the motion search information points to the enlarged reference image. Starting from the position, a block having a size specified according to block division information (block shape information obtained by further dividing a macroblock) included in the motion search information output from the luminance change motion detector 22 is an integer. Obtained as a reference block by sub-sampling in units of precision, and pointed to motion vectors By adding or subtracting a predetermined offset value that is set in advance for a pixel having a phase from the integer pixel positions location, and carries out a process of outputting the image of the reference block as adaptive prediction image.
The luminance change motion detection unit 22 outputs the motion search information to the adaptive prediction image generation unit 21 while appropriately changing the motion search information in the search range, and the motion search information output from the adaptive prediction image generation unit 21. And an error between the adaptive prediction image corresponding to the image and the macroblock image output from the image dividing unit 1, and in the search range, motion search information when the evaluation is optimal is output as inter-frame prediction information, A process of outputting an adaptive prediction image when the evaluation is optimum as an inter-frame prediction image is performed.

FIG. 3 is a block diagram showing a moving picture decoding apparatus according to Embodiment 1 of the present invention.
In FIG. 3, an entropy decoding unit 31 entropy-decodes the encoded data output from the moving picture encoding apparatus of FIG. 1 to obtain quantization transform coefficients, interframe prediction information, intraframe prediction information, and macroblock mode information. Perform the output process. The entropy decoding unit 31 constitutes a decoding unit.

The inverse quantization transform unit 32 performs the same operation as the inverse quantization transform unit 4 in the moving image coding apparatus in FIG. 1, and performs inverse prediction by quantizing the quantized transform coefficient output from the entropy decoding unit 31. A process of outputting an error signal is performed.
The adding unit 33 adds the decoded prediction error signal output from the inverse quantization conversion unit 32 and the predicted image output from the predicted image selection switch 37, and performs a process of outputting a decoded image as a result of the addition. .
The inverse quantization conversion unit 32 and the addition unit 33 constitute decoded image output means.

The frame memory 34 is a memory that stores the decoded image output from the adding unit 33.
When the interframe prediction information is output from the entropy decoding unit 31, the luminance change motion compensation unit 35 reads the decoded image stored in the frame memory 34 as an interframe reference image, and performs interframe prediction according to the interframe prediction information. By performing inter-frame prediction using a reference image, a process of generating and outputting an inter-frame prediction image is performed.
That is, the luminance change motion compensation unit 35 performs adaptive prediction according to the interframe prediction information, as in the luminance change motion compensation prediction unit 7 of the video decoding device, and is included in the interframe prediction information. By interpolating the inter-frame reference image specified by the reference frame number, an expanded reference image in decimal precision unit is generated, and the expanded reference image is then generated according to the block division information included in the inter-frame prediction information. With respect to the pixel value in the reference block obtained by sub-sampling in integer precision units starting from the position indicated by the motion vector included in the inter-frame prediction information, the pixel having a specific phase with respect to the integer component of the motion vector On the other hand, a predetermined offset value set in advance is added or subtracted, and the image of the reference block is output as an inter-frame prediction image.
The luminance change motion compensation unit 35 constitutes a predicted image generation unit.

When intra-frame prediction information is output from the entropy decoding unit 31, the intra-predicted image generation unit 36 reads out the decoded image stored in the frame memory 34 as an intra-frame reference image, and refers to the intra-frame according to the intra-frame prediction information. By performing intra-frame prediction using an image, a process of generating and outputting an intra-frame prediction image is performed.
If the macroblock mode information output from the entropy decoding unit 31 indicates that the optimal macroblock prediction mode is the intraframe prediction mode, the prediction image selection switch 37 outputs from the intra prediction image generation unit 36. If an intra-frame prediction image is selected and the macroblock mode information indicates that the optimal macroblock prediction mode is the interframe prediction mode, the interframe prediction image output from the luminance change motion compensation unit 35 is selected. A process of selecting and outputting the selected image as a predicted image is performed.

Next, the operation will be described.
FIG. 4 is a flowchart showing the processing contents of the moving picture coding apparatus according to the first embodiment of the present invention, and FIG. 5 is a flowchart showing the processing contents of the moving picture decoding apparatus according to the first embodiment of the present invention.
First, the processing content of the moving picture coding apparatus will be described.

When the moving image signal of the moving image is input in frame units or field units, the image dividing unit 1 divides each frame constituting the moving image into macro blocks of a predetermined size, and the macro block which is an image of the macro block An image is output (step ST1).
When the subtracting unit 2 receives the macroblock image from the image dividing unit 1, the subtracting unit 2 obtains a difference between the macroblock image and a predicted image output from the predicted image switching switch 9 described later, and generates a prediction error signal indicating the difference. Output (step ST2).

Upon receiving the prediction error signal from the subtraction unit 2, the quantization conversion unit 3 performs conversion processing and quantization processing on the prediction error signal, and outputs a quantization conversion coefficient of the prediction error signal (step ST3). .
Note that, as described above, the transformation processing includes, for example, orthogonal frequency transformation such as DCT or FFT, or orthogonal transformation such as wavelet transformation.
Further, as described above, for example, the quantization process corresponds to a process such as scalar quantization or vector quantization.

  Upon receiving the quantized transform coefficient from the quantized transform unit 3, the inverse quantization transform unit 4 performs inverse quantization and inverse transform on the quantized transform coefficient to obtain a decoded prediction error signal, and outputs the decoded prediction error signal (Step ST4).

When the luminance change motion compensation prediction unit 7 receives the macroblock image from the image dividing unit 1, the luminance change motion compensation prediction unit 7 reads out the local decoded image stored in the frame memory 6 as an interframe prediction reference image, and the interframe prediction reference image and the macroblock. An inter-frame prediction image is generated by performing inter-frame prediction using an image.
Hereinafter, the processing content of the luminance change motion compensation prediction unit 7 will be specifically described.

First, the adaptive prediction image generation unit 21 of the luminance change motion compensation prediction unit 7 frames a reference image specified by the reference frame number included in the motion search information among the locally decoded images stored in the frame memory 6. By reading out as an inter prediction reference image (an image in units of integer pixels) and interpolating the inter frame prediction reference image (for example, interpolation by interpolation), an enlarged reference image in decimal precision units is generated.
Here, FIG. 6 is an explanatory diagram illustrating an example of generation of an enlarged reference image with a quarter-pixel accuracy.

Next, the adaptive predicted image generation unit 21 uses the position indicated by the motion vector (vector referred to in the motion search) included in the motion search information as the motion search information output from the luminance change motion detection unit 22. A block having a specified size is obtained as a reference block by performing sub-sampling in units of integer precision in accordance with the block division information contained therein (block shape information obtained by further dividing a macroblock).
FIG. 6 shows an example of a reference image enlarged four times in the horizontal and vertical directions.
In FIG. 6, a black circle pixel is a pixel at a position (integer pixel position) indicated by an integer component of a motion vector (vector referred to in motion search) included in the motion search information.
A pixel indicated by a vertical striped circle is a pixel having a specific phase (a phase of a quarter pixel in the lower right direction) with respect to a pixel at an integer pixel position, and a predetermined offset value is added thereto. This is a target pixel (hereinafter referred to as “brightness change addition pixel”).
In addition, the pixels indicated by the horizontal striped circles are pixels having a specific phase (phase of three quarters of pixels in the lower right direction) with respect to the pixels at the integer pixel positions, and a predetermined offset value is subtracted. The target pixel (hereinafter referred to as “luminance change subtraction pixel”).

  That is, the reference block is obtained by sub-sampling pixels in the integer pixel unit with respect to the pixel at the integer pixel position in FIG. For example, when the horizontal component of the motion vector is 1.25 and the vertical component is 3.25, the minority components are each 0.25, so the pixels indicated by the vertical stripe pattern in FIG. A block is constructed.

Next, the adaptive prediction image generation unit 21 calculates the flatness of the reference block. Since the flatness calculation method is a known technique, a description thereof will be omitted.
When the texture of the image is flat, the luminance value at the neighboring pixel position does not change much in the motion search. Therefore, the choice of the predicted image in the motion search is expanded by adding a luminance change to the pixels of some phases. As a result, it is assumed that the probability that the motion prediction is hit increases and the coding efficiency is improved.

Therefore, the adaptive prediction image generation unit 21 refers to the reference block when the flatness of the reference block is larger than a preset threshold value and the reference block is a pixel obtained by sub-sampling the luminance change addition pixel. A predetermined offset value is added to each pixel value of the block. Further, when the flatness of the reference block is larger than a preset threshold and the reference block is a pixel obtained by subsampling each luminance change subtraction pixel, a predetermined offset from each pixel value of the reference block Subtract the value.
Conversely, if the flatness of the block is not greater than a preset threshold value, the offset value is not added to or subtracted from each pixel value of the reference block.
When the adaptive prediction image generation unit 21 completes the offset value addition / subtraction processing, the adaptive prediction image generation unit 21 outputs the image of the block to the luminance change motion detection unit 22 as an adaptive prediction image (step ST5).

The luminance change motion detection unit 22 outputs the motion search information to the adaptive prediction image generation unit 21 while appropriately changing the motion search information in the search range, and the adaptive prediction image generation unit 21 converts the motion search information into the motion search information. When a corresponding adaptive prediction image is received, an error between the adaptive prediction image and the macroblock image output from the image dividing unit 1 is evaluated (step ST6).
The luminance change motion detection unit 22 outputs motion search information when the evaluation is optimal in the search range to the entropy encoding unit 11 as inter-frame prediction information and adaptive prediction when the evaluation is optimal. The image is output to the predicted image changeover switch 9 as an inter-frame predicted image (step ST7).

When receiving the macroblock image from the image dividing unit 1, the intra prediction unit 8 performs intra-frame prediction, which is intra prediction search, in parallel with the process of the luminance change motion compensation prediction unit 7, thereby obtaining an intra-frame prediction image. Generate (step ST8).
That is, the intra prediction unit 8 reads a locally decoded image stored in the frame memory 6 as an intra-frame prediction reference image, and uses the intra-frame prediction reference image to output the macroblock image output from the image division unit 1 and The prediction mode when the evaluation value of the difference between the two is minimized is output to the entropy encoding unit 11 as intra-frame prediction information, and a prediction image generated by the intra prediction method specified by the intra-frame prediction information is output within the frame. The predicted image is output to the predicted image switch 9 as a predicted image.

For example, the encoding control unit 10 determines an optimal macroblock prediction mode and a quantization parameter based on the prediction results of the luminance change motion compensation prediction unit 7 and the intra prediction unit 8 (step ST9), and selects the selection target. Macroblock prediction mode information indicating an image (an intra-frame prediction image or an inter-frame prediction image) is output to the prediction image switch 9.
Also, the encoding control unit 10 outputs an optimal macroblock prediction mode and quantization parameter to the entropy encoding unit 11 as macroblock mode information.
Any method may be used for determining the optimum macroblock prediction mode, and it is sufficient to use a known technique. Therefore, detailed description thereof is omitted here.

  When the prediction image changeover switch 9 receives the macroblock prediction mode information from the encoding control unit 10, the prediction image changeover switch 9 outputs the interframe prediction image output from the luminance change motion compensation prediction unit 7 or the intra according to the macroblock prediction mode information. One of the intra-frame prediction images output from the prediction unit 8 is selected, and the selected prediction image is output to the subtraction unit 2 and the addition unit 5.

  When the addition unit 5 receives the decoded prediction error signal from the inverse quantization conversion unit 4 and receives the prediction image from the prediction image change-over switch 9, the addition unit 5 adds the decoded prediction error signal and the prediction image, and a local part that is the result of the addition The decoded image is stored in the frame memory 6 (step ST10). At this time, the clipping process may be performed so that the pixel value of the locally decoded image obtained as a result of the addition falls within a predetermined range.

The entropy encoding unit 11 includes the quantization transform coefficient output from the quantization transform unit 3, the interframe prediction information output from the luminance change motion compensation prediction unit 7, and the intraframe prediction output from the intra prediction unit 8. The encoded data is output by compressing the information and the macroblock mode information output from the encoding control unit 10 using a predetermined entropy encoding method (step ST11).
The processes in steps ST1 to ST11 described above are repeated until all the macroblocks in the frame are performed, and the process for one frame is completed.

Next, processing contents of the video decoding device will be described.
When the encoded data output from the moving image encoding apparatus in FIG. 1 is input, the entropy decoding unit 31 entropy decodes the encoded data and outputs the quantized transform coefficient to the inverse quantization transform unit 32. The inter-frame prediction information is output to the luminance change motion compensation unit 35, the intra-frame prediction information is output to the intra prediction image generation unit 36, and the macroblock mode information is output to the prediction image selection switch 37 (step ST21).
When receiving the quantization transform coefficient from the entropy decoding unit 31, the inverse quantization transform unit 32 performs inverse quantization on the quantized transform coefficient in the same manner as the inverse quantization transform unit 4 in the moving image coding apparatus in FIG. The decoded prediction error signal is output to the adding unit 33 (step ST22).
In addition, since the quantization parameter used as the unit of the inverse quantization at the time of performing an inverse quantization process is contained in the inter-frame prediction information, it is acquired from the inter-frame prediction information.

When the intra-prediction information is output from the entropy decoding unit 31, the intra-predicted image generation unit 36 generates an intra-frame prediction image (step ST23).
That is, when receiving the intra-frame prediction information from the entropy decoding unit 31, the intra-predicted image generation unit 36 reads the decoded image stored in the frame memory 34 as an intra-frame reference image, and performs intra-frame prediction according to the intra-frame prediction information. By performing intra-frame prediction using the reference image, an intra-frame prediction image is generated and output to the predicted image selection switch 37.

When the interframe prediction information is output from the entropy decoding unit 31, the luminance change motion compensation unit 35 reads the decoded image stored in the frame memory 34 as an interframe reference image, and in accordance with the interframe prediction information, An inter-frame prediction image is generated by performing inter-frame prediction using the inter-reference image (step ST24).
That is, when the luminance change motion compensation unit 35 receives the inter-frame prediction information from the entropy decoding unit 31, the luminance change motion compensation unit 35 adapts according to the inter-frame prediction information in the same manner as the luminance change motion compensation prediction unit 7 of the moving image encoding device in FIG. Predictive.
Hereinafter, the processing content of the luminance change motion compensation unit 35 will be specifically described.

First, similarly to the adaptive prediction image generation unit 21 of the video decoding device, the luminance change motion compensation unit 35 includes the reference frame number included in the interframe prediction information among the decoded images stored in the frame memory 34. The reference image specified by is read as an inter-frame reference image (an image in units of integer pixels), and the inter-frame reference image is interpolated (for example, interpolated by interpolation) to generate an enlarged reference image in decimal precision units. .
Next, the luminance change motion compensation unit 35 converts the block of the specified size into an integer precision unit according to the block division information included in the interframe prediction information (= motion search information) output from the entropy decoding unit 31. To obtain a reference block by performing sub-sampling (see FIG. 6).

Next, the luminance change motion compensation unit 35 calculates the flatness of the reference block.
Similar to the adaptive prediction image generation unit 21 of the video decoding device, the luminance change motion compensation unit 35 has a flatness of the block larger than a preset threshold value, and the reference block subtracts the luminance change addition pixel. When the pixel is obtained by sampling, a predetermined offset value is added to each pixel value of the reference block. Further, when the flatness of the reference block is larger than a preset threshold value and the reference block is a pixel obtained by sub-sampling the luminance change subtraction pixel, a predetermined offset value from each pixel value of the reference block Is subtracted.
Conversely, if the flatness of the block is not greater than a preset threshold value, the offset value is not added to or subtracted from each pixel value of the reference block.
After completing the offset value addition / subtraction process, the luminance change motion compensation unit 35 outputs the image of the block to the predicted image selection switch 37 as an inter-frame predicted image.

  If the macroblock mode information output from the entropy decoding unit 31 indicates that the optimal macroblock prediction mode is the intra-frame prediction mode, the predicted image selection switch 37 is output from the intra predicted image generation unit 36. If the intra-frame prediction image is selected and the macroblock mode information indicates that the optimum macroblock prediction mode is the interframe prediction mode, the interframe prediction image output from the luminance change motion compensation unit 35 is selected. Is selected, and the selected image is output to the adding unit 33 as a predicted image.

  When the addition unit 33 receives the decoded prediction error signal from the inverse quantization conversion unit 32 and receives the prediction image from the prediction image selection switch 37, the addition unit 33 adds the decoded prediction error signal and the prediction image, and is the result of the addition. The decoded image is output to the outside, and the decoded image is stored in the frame memory 34 (step ST25).

  As apparent from the above, according to the first embodiment, the luminance change motion compensation prediction unit 7 interpolates the inter-frame prediction image, which is a locally decoded image stored in the frame memory 6, so that the decimal precision is improved. For the pixel value in the reference block obtained by generating a unit expanded reference image and sub-sampling in integer precision units starting from the position indicated by the motion vector referenced by motion search for the expanded reference image, An adaptive prediction image generation unit 21 that adds / subtracts a predetermined offset value to / from a pixel value of a pixel having a specific phase with respect to an integer component of a motion vector, and outputs an image of the reference block as an adaptive prediction image; The error between the adaptive prediction image output from the generation unit 21 and the macroblock image output from the image division unit 1 is evaluated, and the motion vector when the evaluation is optimized is evaluated. And the brightness change motion detection unit 22 that outputs an adaptive prediction image when the evaluation is optimized as a prediction image, and the brightness of the image changes. At the same time, even when a natural image in which a subject moves is encoded, there is an effect that high-efficiency encoding can be realized without causing deterioration in image quality.

  Further, according to the first embodiment, the luminance change motion compensation unit 35 generates an enlarged reference image in decimal precision units by interpolating an inter-frame reference image that is a decoded image stored in the frame memory 34. In the above reference block obtained by sub-sampling the enlarged reference image in integer precision units starting from the position indicated by the motion vector included in the inter-frame prediction information output from the entropy decoding unit 31 Since a predetermined offset value set in advance for a pixel having a specific phase with respect to the integer component of the motion vector is added to or subtracted from the pixel value, and the image of the reference block is output as a predicted image. There is an effect that a moving image encoded with high efficiency can be decoded.

In the first embodiment, among the decimal precision pixels, a pixel having a phase of a quarter pixel in the lower right direction with respect to the integer pixel position is set as the luminance change addition pixel, and the lower right direction with respect to the integer pixel position. In the above example, a pixel having a phase of 3/4 pixels is used as a luminance change subtraction pixel. However, the pixel does not necessarily have to have this phase, and has a predetermined phase including overlapping with an integer pixel position. The pixel may be a luminance change addition pixel and a luminance change subtraction pixel.
Also, an example has been shown in which one pixel having a predetermined phase is used as a luminance change addition pixel and one other pixel is used as a luminance change subtraction pixel for one integer pixel position. It may be a luminance change addition pixel or a luminance change subtraction pixel.

In the first embodiment, the adaptive prediction image generation unit 21 and the luminance change motion compensation unit 35 of the luminance change motion compensation prediction unit 7 have been described as sharing a preset threshold value. The converting apparatus may transmit the threshold value to the video decoding apparatus.
In this case, for example, the moving image encoding apparatus can be configured to encode and transmit the threshold value by assigning a variable length code such as a fixed length code or a Golomb code in the sequence header (see FIG. 7). reference). Of course, you may transmit using a picture header, a slice header, user data, or auxiliary information data.
In the moving picture decoding apparatus, the threshold value can be acquired by decoding the encoded data of the threshold value included in the header information.
As for the threshold value, the flatness of the entire screen is calculated, and when there are many flat parts, the threshold value for the flatness is set low, thereby increasing the probability that the offset value is added or subtracted, thereby increasing the prediction efficiency. Can do.
In this Embodiment 1, although demonstrated including intra prediction, the presence or absence of intra prediction is not directly related to this invention.

Embodiment 2. FIG.
In the first embodiment, the adaptive predicted image generation unit 21 and the luminance change motion compensation unit 35 of the luminance change motion compensation prediction unit 7 calculate the flatness of the reference block, and the flatness is preset. If it is larger than the threshold value, the offset value is added to or subtracted from the pixel value of the pixel having a specific phase, and if the flatness is not larger than the preset threshold value, the offset value is not added or subtracted. If the adaptive predicted image generation unit 21 and the luminance change motion compensation unit 35 of the luminance change motion compensation prediction unit 7 have a specific phase if the size of the block specified by the block division information is larger than a preset threshold value. Add / subtract the offset value to / from the pixel value of the pixel, and if the block size is not larger than the preset threshold value, add / subtract the offset value. May also be not adversely, the same effects as in the first embodiment.

If the block size to divide the macroblock is the same size as the macroblock, the image is considered to be relatively stable, so the prediction efficiency is improved by adding and subtracting the offset value to expand the options for the predicted image It is thought that it can be made.
Conversely, when the block size is small, the motion and texture are complex, and it is considered that the prediction efficiency can be improved by accurately predicting the motion.
Therefore, in the second embodiment, when the block size is larger than a preset threshold value for the luminance change addition pixel and the luminance change subtraction pixel, the offset value is added to or subtracted from the pixel value. When the size is not larger than a preset threshold value, control is performed so that the offset value is not added to or subtracted from the pixel value.

Embodiment 3 FIG.
In the first embodiment, the adaptive predicted image generation unit 21 and the luminance change motion compensation unit 35 of the luminance change motion compensation prediction unit 7 calculate the flatness of the reference block, and the flatness is preset. If it is larger than the threshold value, the offset value is added to or subtracted from the pixel value of the pixel having a specific phase, and if the flatness is not larger than the preset threshold value, the offset value is not added or subtracted. The adaptive prediction image generation unit 21 and the luminance change motion compensation unit 35 of the luminance change motion compensation prediction unit 7 perform the quantization parameter (inverse) when the quantization process is performed by the quantization conversion unit 3. If the quantization parameter, which is a unit of inverse quantization when the inverse quantization process is performed by the quantization conversion unit 32, is larger than a preset threshold value, the pixel having a specific phase If the offset value is added to or subtracted from the prime value and the quantization parameter is not greater than a preset threshold value, the offset value may not be added or subtracted, and the same effect as in the first embodiment is achieved. .

When the quantization parameter is large, the motion cannot be accurately captured, and it is considered that the prediction efficiency can be improved by adding / subtracting the offset value in order to widen the options of the predicted image.
On the other hand, when the quantization parameter is small, the motion can be accurately grasped. Therefore, it is considered that the prediction efficiency can be improved by not performing the addition / subtraction of the offset value.
Therefore, in the third embodiment, when the quantization parameter is larger than a preset threshold value for the luminance change addition pixel and the luminance change subtraction pixel, addition and subtraction of the offset value with respect to the pixel value is performed. When the quantization parameter is not larger than a preset threshold value, control is performed so that the offset value is not added to or subtracted from the pixel value.

Embodiment 4 FIG.
In the first embodiment, the adaptive predicted image generation unit 21 and the luminance change motion compensation unit 35 of the luminance change motion compensation prediction unit 7 calculate the flatness of the reference block, and the flatness is preset. If it is larger than the threshold value, the offset value is added to or subtracted from the pixel value of the pixel having a specific phase, and if the flatness is not larger than the preset threshold value, the offset value is not added or subtracted. The adaptive prediction image generation unit 21 and the luminance change motion compensation unit 35 of the luminance change motion compensation prediction unit 7 calculate the vector length of the motion vector, and if the vector length is larger than a preset threshold, a specific phase Addition / subtraction of the offset value to / from the pixel value of the pixel having, and if the vector length is not greater than a preset threshold value, the addition / subtraction of the offset value is not performed Unishi at best, the same effects as in the first embodiment.

When the motion vector is large, it is a fast motion and the target object is blurred and the motion cannot be accurately captured, so the prediction efficiency is improved by adding and subtracting the offset value to widen the options of the predicted image It is considered possible.
On the other hand, when the motion vector is small, the motion is small and the target object is clear and it is easy to accurately capture the motion, so that the prediction efficiency can be improved by not adding or subtracting the offset value. it is conceivable that.
Therefore, in the fourth embodiment, when the motion vector is larger than the preset threshold value for the luminance change addition pixel and the luminance change subtraction pixel, addition / subtraction of the offset value is performed on the pixel value, and conversely, When the vector is not large, control is performed so as not to add / subtract the offset value with respect to the pixel value.

Embodiment 5 FIG.
In the first embodiment, the case where only one threshold value used by the adaptive prediction image generation unit 21 and the luminance change motion compensation unit 35 of the luminance change motion compensation prediction unit 7 is set is shown. When the threshold is set, the adaptive prediction image generation unit 21 and the luminance change motion compensation unit 35 of the luminance change motion compensation prediction unit 7 set different offset values according to the comparison result with the plurality of thresholds. You may make it add and subtract.

That is, a plurality of threshold values used by the adaptive prediction image generation unit 21 and the luminance change motion compensation unit 35 of the luminance change motion compensation prediction unit 7 are provided, and the flatness (or block size, quantization parameter, vector length of the motion vector) is provided. ) And a plurality of threshold values, and offset values having different sizes may be added or subtracted depending on the comparison result.
For example, when the threshold value Th1> Th2 and the flatness P is P> Th1, “2” is added to or subtracted from the luminance change addition pixel and the luminance change subtraction pixel, and when Th1>P> Th2. Then, “1” is added to or subtracted from the luminance change addition pixel and the luminance change subtraction pixel, and when Th2> P, control is performed so as not to add or subtract the offset value. The block size, the quantization parameter, and the vector length of the motion vector are similarly controlled.

Embodiment 6 FIG.
In the first embodiment, the case where only one threshold value used by the adaptive prediction image generation unit 21 and the luminance change motion compensation unit 35 of the luminance change motion compensation prediction unit 7 is set is shown. When the threshold value is set, the adaptive prediction image generation unit 21 and the luminance change motion compensation unit 35 of the luminance change motion compensation prediction unit 7 perform specific addition / subtraction of the offset value according to the comparison result with a plurality of threshold values. The number of pixels having a phase may be changed.

That is, a plurality of threshold values used by the adaptive prediction image generation unit 21 and the luminance change motion compensation unit 35 of the luminance change motion compensation prediction unit 7 are provided, and the flatness (or block size, quantization parameter, vector length of the motion vector) is provided. ) And a plurality of threshold values, and the number of luminance change addition pixels and luminance change subtraction pixels may be changed according to the comparison result.
For example, when the threshold Th1> Th2, when the flatness P is P> Th1, two luminance change addition pixels and two luminance change subtraction pixels are provided, and when Th1>P> Th2, the luminance is obtained. When Th2> P, one change addition pixel and one luminance change subtraction pixel are controlled so as not to add / subtract the offset value. The block size, the quantization parameter, and the vector length of the motion vector are similarly controlled.

  The moving image encoding device, the moving image decoding device, and the like according to the present invention provide high-efficiency encoding without degrading image quality even when encoding a natural image in which a subject moves at the same time as the brightness of the image changes. Therefore, it is suitable for use in a moving image encoding device that encodes moving images with high efficiency and a moving image decoding device that decodes moving images encoded with high efficiency. .

Claims (16)

  An image dividing unit that divides each frame constituting a moving image into macroblocks of a predetermined size and outputs an image of the macroblock; a difference between a macroblock image that is an image output from the image dividing unit and a predicted image Quantization conversion means for performing a conversion process and a quantization process on the prediction error signal indicating the difference and outputting a quantization conversion coefficient of the prediction error signal, and a quantum output from the quantization conversion means Decoding prediction error signal is obtained by inverse quantization and inverse transformation of the generalized transform coefficient, the decoded prediction error signal and the prediction image are added, and a local decoded image output means for outputting a local decoded image as a result of the addition And inter-frame prediction using the macroblock image output from the image dividing unit and the local decoded image output from the local decoded image output unit The inter-frame prediction means for generating the predicted image and outputting inter-frame prediction information for specifying the inter-frame prediction method, the quantized transform coefficient output from the quantizing transform means, and the inter-frame In the moving picture coding apparatus comprising: encoding means for entropy encoding the interframe prediction information output from the prediction means and outputting encoded data that is a result of the encoding, the interframe prediction means includes: By interpolating the local decoded image output from the local decoded image output means, an enlarged reference image in decimal precision unit is generated, and an integer starting from the position indicated by the motion vector referred to by the motion search with respect to the enlarged reference image It has a specific phase for the integer component of the motion vector for the pixel value in the reference block obtained by sub-sampling with accuracy unit An adaptive prediction image generation unit that adds and subtracts a predetermined offset value set in advance to pixels and outputs the image of the reference block as an adaptive prediction image, and an adaptive output output from the adaptive prediction image generation unit An error between the predicted image and the macroblock image output from the image dividing unit is evaluated, and a motion vector when the evaluation is optimized is output as the inter-frame prediction information, and adaptation when the evaluation is optimized A moving image encoding apparatus comprising: a luminance change motion detecting unit that outputs a target predicted image as the predicted image.   The adaptive prediction image generation unit calculates the flatness of the reference block, and if the flatness is larger than a preset threshold value, the adaptive prediction image generation unit offsets the pixel value of the pixel having a specific phase with respect to the integer component of the motion vector. 2. The moving picture coding apparatus according to claim 1, wherein the offset value is not added or subtracted unless the flatness is larger than a preset threshold value.   If the size of the reference block is larger than a preset threshold, the adaptive prediction image generation unit adds or subtracts an offset value to or from a pixel value of a pixel having a specific phase with respect to the integer component of the motion vector, and the size of the block 2. The moving picture encoding apparatus according to claim 1, wherein the offset value is not added or subtracted if is not greater than a preset threshold value.   The adaptive prediction image generation unit, if a quantization parameter that is a unit of quantization when the quantization process is performed by the quantization conversion unit is greater than a preset threshold value, 2. The moving image according to claim 1, wherein an offset value is added to or subtracted from a pixel value of a pixel having a phase, and the offset value is not added or subtracted unless the quantization parameter is larger than a preset threshold value. Encoding device.   The adaptive prediction image generation unit calculates a vector length of the motion vector, and if the vector length is larger than a preset threshold, an offset value is added to the pixel value of the pixel having a specific phase with respect to the integer component of the motion vector. 2. The moving picture encoding apparatus according to claim 1, wherein addition / subtraction is not performed if the addition / subtraction is performed and the vector length is not greater than a preset threshold value.   3. The moving image according to claim 2, wherein, when a plurality of threshold values are set, the adaptive prediction image generation unit adds or subtracts offset values having different sizes in accordance with a comparison result with the plurality of threshold values. Encoding device.   When a plurality of threshold values are set, the adaptive prediction image generation unit changes the number of pixels having a specific phase with respect to the integer component of the motion vector to which the offset value is added or subtracted according to the comparison result with the plurality of threshold values. The moving picture coding apparatus according to claim 2, wherein:   Decoding means for entropy decoding the encoded data and outputting quantized transform coefficients and inter-frame prediction information; and dequantizing and inverse transforming the quantized transform coefficients output from the decoding means to generate a decoded prediction error signal A decoded image output means for adding the decoded prediction error signal and the predicted image and outputting a decoded image as a result of the addition; and the decoded image output means according to the interframe prediction information output from the decoding means. In the moving picture decoding apparatus including the predicted image generation unit that generates the predicted image by performing inter-frame prediction using the decoded image output from the decoded image, the predicted image generation unit includes the decoded image output unit By interpolating the decoded image output from, an enlarged reference image in decimal precision unit is generated, and the inter-frame prediction information for the enlarged reference image For a pixel value in the reference block obtained by subsampling in integer precision units starting from the position indicated by the vector, a predetermined value set in advance for a pixel having a specific phase with respect to the integer component of the motion vector A moving picture decoding apparatus characterized by adding and subtracting an offset value and outputting an image of the reference block as the predicted image.   The predicted image generation means calculates the flatness of the reference block, and if the flatness is larger than a preset threshold value, the predicted image generation means sets an offset value to the pixel value of the pixel having a specific phase with respect to the integer component of the motion vector. 9. The moving picture decoding apparatus according to claim 8, wherein addition / subtraction is performed, and addition / subtraction of the offset value is not performed unless the flatness is larger than a preset threshold value.   If the size of the reference block is larger than a preset threshold, the predicted image generation means adds or subtracts an offset value to or from a pixel value of a pixel having a specific phase with respect to the integer component of the motion vector, and the size of the block is determined in advance. 9. The moving picture decoding apparatus according to claim 8, wherein addition / subtraction of the offset value is not performed unless the threshold value is larger than a set threshold value.   The predicted image generation unit is configured to specify a specific value for an integer component of a motion vector if a quantization parameter that is a unit of dequantization when the inverse quantization process is performed by the decoded image output unit is greater than a preset threshold value. 9. The moving image according to claim 8, wherein an offset value is added to or subtracted from a pixel value of a pixel having a phase, and the offset value is not added or subtracted unless the quantization parameter is larger than a preset threshold value. Decoding device.   The predicted image generation means calculates the vector length of the motion vector, and if the vector length is greater than a preset threshold value, adds or subtracts an offset value to the pixel value of a pixel having a specific phase with respect to the integer component of the motion vector. 9. The moving picture decoding apparatus according to claim 8, wherein the offset value is not added or subtracted unless the vector length is greater than a preset threshold value.   10. The moving picture decoding apparatus according to claim 9, wherein, when a plurality of threshold values are set, the predicted image generation unit adds or subtracts offset values having different sizes in accordance with a comparison result with the plurality of threshold values. .   When a plurality of threshold values are set, the predicted image generation means changes the number of pixels having a specific phase with respect to the integer component of the motion vector to which the offset value is added or subtracted according to the comparison result with the plurality of threshold values. The moving picture decoding apparatus according to claim 9, wherein:   The image dividing means divides each frame constituting the moving image into macroblocks of a predetermined size, and outputs an image of the macroblock. The quantization transforming means is an image output in the image dividing step. A quantization transform step for obtaining a difference between a certain macroblock image and a predicted image, performing a transform process and a quantization process on a prediction error signal indicating the difference, and outputting a quantized transform coefficient of the prediction error signal; The decoded image output means dequantizes and inversely transforms the quantized transform coefficient output in the quantization transform step to obtain a decoded prediction error signal, adds the decoded prediction error signal and the predicted image, and adds A local decoded image output step for outputting a local decoded image as a result of the above, and a macroblock output by the inter-frame prediction means in the image division step Inter-frame prediction is performed by performing inter-frame prediction using the local decoded image output in the local decoded image output step and inter-frame prediction information for specifying the inter-frame prediction method. The output interframe prediction step, the encoding means entropy encodes the quantized transform coefficient output in the quantization transform step and the interframe prediction information output in the interframe prediction step, An encoding step for outputting certain encoded data, wherein the inter-frame prediction means interpolates the local decoded image output in the local decoded image output step, thereby obtaining a decimal precision unit. The enlarged reference image is generated, and the position indicated by the motion vector referred to by the motion search for the enlarged reference image is adjusted as the starting point. The pixel value in the reference block obtained by sub-sampling in accuracy units is added or subtracted with a predetermined offset value set in advance to the pixel having a specific phase with respect to the integer component of the motion vector, and the image of the reference block is An adaptive prediction image generation step that outputs as an adaptive prediction image; an error between the adaptive prediction image output in the adaptive prediction image generation step and the macroblock image output in the image division step; And a luminance change motion detection step of outputting an adaptive prediction image when the evaluation is optimized as the prediction image and outputting a motion vector when the evaluation is optimal as the inter-frame prediction information. A video encoding method.   A decoding step in which the decoding means entropy-decodes the encoded data and outputs quantized transform coefficients and inter-frame prediction information; and a decoded image output means dequantizes and reverses the quantized transform coefficients output in the decoding step. A decoded prediction error signal is converted to obtain a decoded prediction error signal, the decoded prediction error signal and the prediction image are added, and a decoded image output step for outputting a decoded image as a result of the addition; In a video decoding method comprising: a predicted image generation step of generating the predicted image by performing inter-frame prediction using the decoded image output in the decoded image output step according to the inter-frame prediction information performed The predicted image generation means interpolates the decoded image output in the decoded image output step, thereby expanding the reference image in decimal precision units. A motion vector is generated with respect to the pixel value in the reference block obtained by sub-sampling in an integer precision unit from the position indicated by the motion vector as the inter-frame prediction information with respect to the enlarged reference image. A moving picture decoding method comprising: adding and subtracting a predetermined offset value set in advance to a pixel value of a pixel having a specific phase with respect to an integer component of the integer component, and outputting an image of the reference block as the predicted image.

Images