JP2012186763A - Video encoding device, video decoding device, video encoding method, and video decoding method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To obtain a video encoding device and a video encoding method that are capable of notifying a video decoding device side of a division state of CUs by signaling with a small number of bits.SOLUTION: A variable length encoding unit 13 performs variable length encoding on a maximum CU division layer number indicating a division layer number of a portion in the deepest layer of a quadtree structure in each LCU, and also performs variable length encoding on a division flag indicating whether a CU other than CUs belonging to the deepest layer is divided into a quadtree by a block division unit 2.

Description

  The present invention relates to a moving image encoding device, a moving image decoding device, a moving image encoding method, and a moving image decoding method used for image compression encoding technology, compressed image data transmission technology, and the like.

For example, MPEG, ITU-TH. In a conventional video encoding method such as the 26x series, each frame of a video signal is equally divided into square blocks called “macroblocks”, and the intra-frame / interframe prediction processing and DCT for the prediction differential signal are performed in units of macroblocks. A method of obtaining a bit stream which is final compressed data by performing entropy coding by performing orthogonal transformation processing such as the above and quantization processing is adopted.
AVC / H. In the method up to H.264, the block size of 16 pixels × 16 lines is adopted as the macroblock size on the luminance signal, but this size is expanded to adapt the size of the prediction target block and the transform block. Thus, a technique for greatly improving the coding performance has been reported (see, for example, Non-Patent Document 1).

  In the conventional macroblock size expansion method disclosed in Non-Patent Document 1, a macroblock is referred to as an LCU (Large Coding Unit), and the LCU is further divided into blocks after division. Have disclosed a configuration for detecting a motion vector and a configuration for adapting the size of orthogonal transform, but a CU (Coding Unit) which is a coding unit when performing intraframe coding or interframe coding is disclosed. The quadtree division shown in FIG. 13 is adopted.

In FIG. 13, an encoded block having a size of (M ⁰ , M ⁰ ) (the number at the upper right shoulder indicates a layer level) having a luminance component expressed as “0th layer” is an LCU. Starting from the LCU, a CU in a divided state is obtained by hierarchically dividing up to the upper limit of the number of hierarchies separately determined for each sequence (upper limit hierarchy number in the sequence) in a quadtree structure.
Since quadtree partitioning is performed, (M ^{n + 1} , M ^{n + 1} ) = (M ⁿ / 2, M ⁿ / 2) always holds.
In the conventional macroblock size expansion method, the division state of the CU is expressed by signaling a division flag indicating whether or not to divide the CU for each CU, starting from the LCU.

For example, as shown in FIG. 14, when an LCU is divided hierarchically in a quadtree structure, one LCU (0th hierarchy), 4 CUs (1st hierarchy), and 12 CUs ( Therefore, a division flag (1 = division: 0 = non-division) indicating whether each of the 17 blocks (LCU + CU) in total is divided is prepared.
For this reason, the moving picture coding apparatus signals 17 division flags in 17 bits in order to notify the moving picture decoding apparatus side of the division state of the CU.

T. Wiegand, W.-J. Han, B. Bross, J.-R. Ohm and GJ Sullivan, “WD1: Working Draft 1 of High-Efficiency Video Coding,” Joint Collaborative Team on Video Coding (JCT-VC) of ITU-T SG16 WP3 and ISO / IEC JTC1 / SC29 / WG11, doc.JCTVC-C403, Guangzhou, China, October 2010.

  Since the conventional moving picture coding apparatus is configured as described above, the division state of the CU having a higher degree of freedom can be expressed as the hierarchy of the quadtree structure is deepened. However, in order to transmit the division state of the CU to the video decoding device side, it is necessary to perform signaling with a large number of bits, and there is a problem that a necessary code amount increases.

The present invention has been made to solve the above-described problems, and is a moving picture coding apparatus and a moving picture coding method capable of transmitting the division state of a CU to the moving picture decoding apparatus side by signaling with a small number of bits. The purpose is to obtain.
Another object of the present invention is to obtain a moving picture decoding apparatus and a moving picture decoding method that can be applied to the above moving picture encoding apparatus.

  The moving picture coding apparatus according to the present invention divides an input image into macroblocks of a predetermined size and further divides each macroblock hierarchically by a quadtree structure, thereby encoding more finely than the macroblock. A block dividing unit that outputs an encoding processing unit block that is a unit block, and a prediction process for the encoding processing unit block in an encoding mode corresponding to the encoding processing unit block output from the block dividing unit A prediction image generating unit that generates a prediction image, a differential image between the encoding processing unit block output from the block dividing unit and the prediction image generated by the prediction image generation unit is compressed, and compressed data of the difference image is output Image compression means, and the compressed data output from the image compression means and the encoding mode are variable-length encoded, and the compressed data And variable-length encoding means for generating a bit stream in which encoded data of the encoding mode and the encoding mode are multiplexed, and the variable-length encoding means is the deepest in the quadtree structure in each macroblock Whether the maximum number of divided hierarchies indicating the number of divided hierarchies is variable-length coded, and whether or not the coding processing unit blocks other than the coding processing unit block belonging to the deepest hierarchy are divided into quadtrees by the block dividing means The division flag indicating the variable length is encoded with variable length.

  According to the present invention, the input image is divided into macroblocks of a predetermined size, and each macroblock is hierarchically divided in a quadtree structure, so that it is a block of a coding unit smaller than the macroblock. A prediction image is generated by performing a prediction process on the encoding processing unit block in a block dividing unit that outputs the encoding processing unit block and an encoding mode corresponding to the encoding processing unit block output from the block dividing unit. A prediction image generation unit; and an image compression unit that compresses a differential image between the encoding processing unit block output from the block division unit and the prediction image generated by the prediction image generation unit, and outputs compressed data of the difference image; Then, the compressed data output from the image compression means and the encoding mode are subjected to variable length encoding, and the compressed data and the encoding mode are encoded. Variable-length encoding means for generating a bit stream in which encoded data of a group is multiplexed, and the variable-length encoding means is a divided hierarchy of the deepest part in the quadtree structure in each macroblock A division flag that indicates whether or not the encoding processing unit block other than the encoding processing unit block belonging to the deepest hierarchy is divided into quadtrees by the block dividing means, while the maximum number of division hierarchies indicating the number is variable-length encoded Therefore, there is an effect that the division state of the encoding processing unit block can be transmitted to the moving picture decoding apparatus side by signaling with a small number of bits.

It is a block diagram which shows the moving image encoder 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 entropy encoding process of a CU division | segmentation state. 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 decoding apparatus by Embodiment 1 of this invention. It is a flowchart which shows the entropy decoding process of a CU division | segmentation state. It is explanatory drawing which shows a mode that LCU is divided | segmented hierarchically and several CU is obtained. (A) shows the distribution of the partitions after the division, and (b) is an explanatory diagram showing a situation where the encoding mode m (B _j ⁿ ) is assigned by the hierarchical division in a quadtree graph. It is explanatory drawing which shows an example of the bit stream produced | generated by the variable length encoding part. It is explanatory drawing which shows the code example of the maximum CU division | segmentation hierarchy number. It is explanatory drawing which shows the division | segmentation flag of 17 blocks (LCU + CU). It is explanatory drawing which shows the example of the Huffman code | cord | chord of the maximum CU division | segmentation hierarchy number. It is explanatory drawing which shows the hierarchical division | segmentation of LCU by a quadtree structure. It is explanatory drawing which shows the division | segmentation flag of 17 blocks (LCU + CU).

Embodiment 1 FIG.
In the first embodiment, each frame image of a video is input, and after performing compression processing by orthogonal transformation or quantization on a prediction difference signal obtained by performing motion compensation prediction between adjacent frames, the variable length is obtained. A video encoding device that performs encoding to generate a bitstream and a video decoding device that decodes a bitstream output from the video encoding device will be described.

The moving picture coding apparatus according to the first embodiment adapts to local changes in the spatial and temporal directions of a video signal, divides the video signal into regions of various sizes, and performs intraframe / interframe adaptive coding. It is characterized by performing.
In general, a video signal has a characteristic that the complexity of the signal changes locally in space and time. When viewed spatially, on a particular video frame, there are patterns with uniform signal characteristics in a relatively large image area such as the sky and walls, and small images such as people and paintings with fine textures. A pattern having a complicated texture pattern may be mixed in the region.
Even when viewed temporally, the sky and the wall have small changes in the pattern in the time direction locally, but the moving person or object has a rigid or non-rigid motion in time, so the temporal change does not occur. large.

The encoding process reduces the overall code amount by generating a prediction difference signal with small signal power and entropy by temporal and spatial prediction, but the parameters for prediction are made uniform in as large an image signal region as possible. If applicable, the code amount of the parameter can be reduced.
On the other hand, if the same prediction parameter is applied to an image signal pattern having a large temporal and spatial change, the number of prediction differential signals cannot be reduced because prediction errors increase.
Therefore, for image signal patterns with large temporal and spatial changes, the prediction target signal power / entropy is reduced even if the prediction target area is reduced and the amount of parameter data for prediction is increased. Is preferable.
In order to perform coding adapted to the general properties of such a video signal, the moving picture coding apparatus of the first embodiment divides the video signal area hierarchically from a predetermined maximum block size, Prediction processing and prediction difference encoding processing are performed for each divided region.

The video signal to be processed by the moving image coding apparatus according to the first embodiment is a color in an arbitrary color space such as a YUV signal composed of a luminance signal and two color difference signals, or an RGB signal output from a digital image sensor. In addition to the video signal, the video frame is an arbitrary video signal such as a monochrome image signal or an infrared image signal, in which the video frame is composed of a horizontal and vertical two-dimensional digital sample (pixel) sequence.
The gradation of each pixel may be 8 bits, or may be gradation such as 10 bits or 12 bits.
However, in the following description, it is assumed that the input video signal is a YUV signal unless otherwise specified. In addition, it is assumed that the two color difference components U and V are subsampled 4: 2: 0 format signals with respect to the luminance component Y.
The processing data unit corresponding to each frame of the video is referred to as “picture”. In the first embodiment, “picture” is described as a signal of a video frame that has been sequentially scanned (progressive scan). However, when the video signal is an interlace signal, the “picture” may be a field image signal which is a unit constituting a video frame.

1 is a block diagram showing a moving picture coding apparatus according to Embodiment 1 of the present invention.
In FIG. 1, the encoding control unit 1 has a size (LCU) of an LCU (macroblock) that is an encoding unit when a motion compensation prediction process (interframe prediction process) or an intra prediction process (intraframe prediction process) is performed. Size), and the upper limit of the number of division layers of a CU (block of encoding unit smaller than LCU) set in advance (the upper limit of the number of division layers of a CU is set, for example, in units of sequences, pictures, or slices) In the quadrant tree structure in each LCU, a process of determining the maximum number of divided CU layers indicating the number of divided layers in the deepest part is performed.
Further, the encoding control unit 1 performs a process of determining an encoding mode suitable for each CU obtained by dividing the LCU until the maximum number of CU division layers is reached. That is, a process of selecting a coding mode suitable for each CU from one or more available coding modes (one or more intra coding modes and one or more inter coding modes) is performed.

  When the block dividing unit 2 inputs a video signal indicating an input image, the block dividing unit 2 divides the input image indicated by the video signal into LCUs having the LCU size determined by the encoding control unit 1 and is determined by the encoding control unit 1. Until the maximum number of CU division hierarchies is reached, the LCU is hierarchically divided by a quadtree structure to obtain a CU, and a process of outputting each CU is performed. The block dividing unit 2 constitutes block dividing means.

If the coding mode selected by the coding control unit 1 is the intra coding mode, the changeover switch 3 outputs the CU divided by the block dividing unit 2 to the intra prediction unit 4 and selects it by the coding control unit 1. If the encoded mode is the inter encoding mode, a process of outputting the CU divided by the block division unit 2 to the motion compensation prediction unit 5 is performed.
When the intra prediction unit 4 receives the CU divided by the block division unit 2 from the changeover switch 3, the intra prediction process is performed on the CU using the intra prediction parameter output from the encoding control unit 1, and the predicted image The process to generate is performed.
When the motion compensation prediction unit 5 receives the CU divided by the block division unit 2 from the changeover switch 3, the motion compensation prediction unit 5 performs the motion compensation prediction process for the CU using the inter prediction parameter output from the encoding control unit 1. A process for generating a predicted image is performed.
The changeover switch 3, the intra prediction unit 4, and the motion compensation prediction unit 5 constitute a predicted image generation unit.

The subtracting unit 6 generates a difference image (= CU-predicted image) by subtracting the prediction image generated by the intra prediction unit 4 or the motion compensation prediction unit 5 from the CU divided by the block dividing unit 2. A process of outputting a prediction difference signal indicating a difference image is performed.
The transform / quantization unit 7 converts the prediction difference signal output from the subtraction unit 6 in units of transform block size included in the prediction difference encoding parameter output from the encoding control unit 1 (for example, DCT ( Discrete cosine transformation) or orthogonal transformation processing such as KL transformation for which a base design is made in advance for a specific learning sequence) and using the quantization parameter included in the prediction differential coding parameter Then, by quantizing the transform coefficient of the prediction difference signal, a process of outputting the quantized transform coefficient as compressed data of the difference image is performed. The transform / quantization unit 7 constitutes an image compression unit.

  The inverse quantization / inverse transform unit 8 performs inverse quantization on the compressed data output from the transform / quantization unit 7 using the quantization parameter included in the prediction difference encoding parameter output from the encoding control unit 1. Inverse transform processing (for example, inverse DCT (Inverse Discrete Cosine Transform), inverse KL transform, etc.) of the inverse quantized compressed data in units of transform block sizes included in the prediction difference encoding parameter ), The process of outputting the compressed data after the inverse transform process as a local decoded prediction difference signal is performed.

The adding unit 9 adds the local decoded prediction difference signal output from the inverse quantization / inverse transform unit 8 and the prediction signal indicating the prediction image generated by the intra prediction unit 4 or the motion compensation prediction unit 5 to thereby perform local decoding. A process of generating a locally decoded image signal indicating an image is performed.
The intra prediction memory 10 is a recording medium such as a RAM that stores a local decoded image indicated by the local decoded image signal generated by the adding unit 9 as an image used in the next intra prediction process by the intra prediction unit 4.

The loop filter unit 11 compensates for the coding distortion included in the locally decoded image signal generated by the adding unit 9, and performs motion compensation prediction using the locally decoded image indicated by the locally decoded image signal after the coding distortion compensation as a reference image. A process of outputting to the frame memory 12 is performed.
The motion compensated prediction frame memory 12 is a recording medium such as a RAM that stores a locally decoded image after the filtering process by the loop filter unit 11 as a reference image used in the next motion compensated prediction process by the motion compensated prediction unit 5.

  The variable length encoding unit 13, for example, the compressed data output from the transform / quantization unit 7, the encoding mode output from the encoding control unit 1, the prediction differential encoding parameter, the maximum number of CU partition layers, and the partition flag ( A flag indicating whether or not a CU other than the CU belonging to the deepest hierarchy has been quadtree-divided by the block dividing unit 2), an intra-prediction parameter output from the intra-prediction unit 4, or a motion-compensated prediction unit 5 The inter-prediction parameter is subjected to variable length coding, and the compressed data, the coding mode, the prediction differential coding parameter, the maximum number of CU partition layers, the partition flag, and the intra-prediction parameter / inter-prediction parameter encoded data are multiplexed. To generate a bitstream. The variable length encoding unit 13 constitutes variable length encoding means.

In FIG. 1, a coding control unit 1, a block division unit 2, a changeover switch 3, an intra prediction unit 4, a motion compensation prediction unit 5, a subtraction unit 6, and a transform / quantization unit 7, which are components of the moving image coding apparatus. , The inverse quantization / inverse transform unit 8, the adder unit 9, the loop filter unit 11 and the variable length coding unit 13 each have dedicated hardware (for example, a semiconductor integrated circuit on which a CPU is mounted, or a one-chip microcomputer) However, when the moving image encoding apparatus is configured by a computer, the encoding control unit 1, the block division unit 2, the changeover switch 3, the intra prediction unit 4, the motion compensation A program describing the processing contents of the prediction unit 5, subtraction unit 6, transformation / quantization unit 7, inverse quantization / inverse transformation unit 8, addition unit 9, loop filter unit 11, and variable length coding unit 13 Computer Stored in the memory, CPU of the computer may execute a program stored in the memory.
FIG. 2 is a flowchart showing the processing contents of the moving picture coding apparatus according to Embodiment 1 of the present invention.
FIG. 3 is a flowchart showing the entropy encoding process in the CU division state.

FIG. 4 is a block diagram showing a moving picture decoding apparatus according to Embodiment 1 of the present invention.
In FIG. 4, the variable length decoding unit 51 includes compressed data, encoding mode, prediction differential encoding parameter, and intra prediction parameter related to each CU divided hierarchically from encoded data multiplexed in the bitstream. / Inter-prediction parameters are variable-length decoded, and the compressed data and prediction differential encoding parameters are output to the inverse quantization / inverse transform unit 55, and the encoding mode and the intra-prediction parameters / inter-prediction parameters are changed over by the switch 52. Execute the process to output to.
In addition, the variable length decoding unit 51 performs variable length decoding on the maximum number of CU division layers and the division flag from the encoded data, and performs processing for decoding the LCU quadtree structure from the maximum number of CU division layers and the division flag. .
The variable length decoding unit 51 constitutes variable length decoding means.

When the coding mode related to the CU output from the variable length decoding unit 51 is the intra coding mode, the changeover switch 52 outputs the intra prediction parameter output from the variable length decoding unit 51 to the intra prediction unit 53, and When the coding mode is the inter coding mode, a process of outputting the inter prediction parameter output from the variable length decoding unit 51 to the motion compensation prediction unit 54 is performed.
The intra prediction unit 53 uses the intra prediction parameter output from the changeover switch 52 to perform an intra prediction process for the CU and generate a predicted image.
The motion compensation prediction unit 54 performs a process of generating a predicted image by performing a motion compensation prediction process for the CU using the inter prediction parameter output from the changeover switch 52.
The changeover switch 52, the intra prediction unit 53, and the motion compensation prediction unit 54 constitute a predicted image generation unit.

The inverse quantization / inverse transform unit 55 uses the quantization parameter included in the prediction difference encoding parameter output from the variable length decoding unit 51 to compress the encoded block output from the variable length decoding unit 51. Data is inversely quantized, and inverse transform processing (for example, inverse DCT (Inverse Discrete Cosine Transform), inverse KL transform, etc.) is performed on the transform block size unit included in the prediction differential encoding parameter. (Inverse transform process) is performed to output the compressed data after the inverse transform process as a decoded prediction difference signal (a signal indicating a difference image before compression).
The addition unit 56 adds the decoded prediction difference signal output from the inverse quantization / inverse conversion unit 55 and the prediction signal indicating the prediction image generated by the intra prediction unit 53 or the motion compensation prediction unit 54, thereby adding the decoded image. The process which produces | generates the decoded image signal shown is implemented.
The inverse quantization / inverse transform unit 55 and the addition unit 56 constitute decoded image generation means.

The intra prediction memory 57 is a recording medium such as a RAM that stores a decoded image indicated by the decoded image signal generated by the adding unit 56 as an image used by the intra prediction unit 53 in the next intra prediction process.
The loop filter unit 58 compensates for the encoding distortion included in the decoded image signal generated by the adding unit 56, and uses the decoded image indicated by the decoded image signal after the encoding distortion compensation as a reference image for the motion compensated prediction frame memory 59. Execute the process to output to.
The motion compensated prediction frame memory 59 is a recording medium such as a RAM that stores a decoded image after the filtering process by the loop filter unit 58 as a reference image to be used by the motion compensation prediction unit 54 in the next motion compensation prediction process.

In FIG. 4, a variable length decoding unit 51, a changeover switch 52, an intra prediction unit 53, a motion compensation prediction unit 54, an inverse quantization / inverse conversion unit 55, an addition unit 56, and a loop filter unit, which are components of the moving picture decoding apparatus. 58 is assumed to be configured by dedicated hardware (for example, a semiconductor integrated circuit on which a CPU is mounted, or a one-chip microcomputer). A program describing the processing contents of the variable length decoding unit 51, the changeover switch 52, the intra prediction unit 53, the motion compensation prediction unit 54, the inverse quantization / inverse transformation unit 55, the addition unit 56, and the loop filter unit 58. May be stored in the memory of the computer, and the CPU of the computer may execute the program stored in the memory.
FIG. 5 is a flowchart showing the processing contents of the moving picture decoding apparatus according to Embodiment 1 of the present invention.
FIG. 6 is a flowchart showing the entropy decoding process in the CU division state.

Next, the operation will be described.
First, the processing contents of the moving picture encoding apparatus in FIG. 1 will be described.
First, the encoding control unit 1 sets the LCU size (LCU size) used for encoding the encoding target picture (current picture) and the upper limit of the number of division hierarchies of the CU set in advance. The maximum number of CU division layers indicating the number of division layers of the deepest part in the quadtree structure in the LCU is determined (step ST1 in FIG. 2).
Also, the encoding control unit 1 determines an encoding mode suitable for each CU obtained by dividing the LCU up to the maximum number of CU division layers by the block dividing unit 2.
That is, a coding mode suitable for each CU is selected from one or more available coding modes (one or more intra coding modes and one or more inter coding modes).
Since the encoding mode selection process is a known technique, a detailed description thereof will be omitted. In the coding mode selection process, after the LCU is hierarchically divided by the block dividing unit 2 to be described later and a plurality of CUs are obtained, a coding mode suitable for each CU may be determined. .

As a method of determining the LCU size, for example, a method of determining a size corresponding to the resolution of the input image for all the pictures can be considered.
Also, a method of quantifying the difference in complexity of local motion of an input image as a parameter, determining a LCU size to a small value for a picture with intense motion, and determining a LCU size to a large value for a picture with little motion And so on.
The maximum number of CU division layers is set so that, for example, the higher the motion of the input image, the deeper the number of layers, and the more detailed motion can be detected. If the motion of the input image is small, the number of layers is suppressed. A method of setting is conceivable.
Note that the maximum number of CU partition layers may be configured to use a fixed value in sequence units.

When the video signal indicating the input image is input, the block dividing unit 2 divides the input image indicated by the video signal into the LCU size determined by the encoding control unit 1.
In addition, the block division unit 2 divides the LCU hierarchically in a quadtree structure until the maximum number of CU division layers determined by the encoding control unit 1 is reached for each LCU size image area (LCU). As a result, CUs, which are blocks of coding units smaller than the LCU, are obtained, and each CU is output to the changeover switch 3 and the subtraction unit 6 (step ST2).

Here, FIG. 7 is an explanatory diagram showing a state in which the LCU is hierarchically divided to obtain a plurality of CUs.
The LCU is defined as a CU having a pixel size of (L ⁰ , M ⁰ ) with a luminance component represented as “0th layer” in FIG.
In the example of FIG. 7, a plurality of CUs are obtained by performing a hierarchical division to a predetermined depth (the maximum number of CU division hierarchies) with a quadtree structure starting from an LCU.

At depth n, the CU is an image area of size (L ⁿ , M ⁿ ).
However, L ⁿ and M ⁿ may be the same or different, but the example of FIG. 7 shows the case of L ⁿ = M ⁿ .
Hereinafter, the size of the CU is defined as the size (L ⁿ , M ⁿ ) in the luminance component of the CU.

Since the block division unit 2 performs quadtree division, (L ^{n + 1} , M ^{n + 1} ) = (L ⁿ / 2, M ⁿ / 2) always holds.
However, in a color video signal (4: 4: 4 format) in which all color components have the same number of samples, such as RGB signals, the size of all color components is (L ⁿ , M ⁿ ). When the 4: 2: 0 format is handled, the size of the corresponding color difference component coding block is (L ⁿ / 2, M ⁿ / 2).
Hereinafter, the n-th layer CU is represented by B _j ⁿ (j: CU number in the n-th layer), and the encoding mode that can be selected by B _j ⁿ is represented by m (B _j ⁿ ).

In the case of a color video signal composed of a plurality of color components, the encoding mode m (B _j ⁿ ) may be configured to use an individual mode for each color component. The description will be made on the assumption that it indicates a coding mode for a luminance component of a YUV signal, 4: 2: 0 format CU.
The coding mode m (B _j ⁿ ) includes one or more intra coding modes (collectively “INTRA”), one or more inter coding modes (collectively “INTER”), As described above, the encoding control unit 1 selects an encoding mode having the highest encoding efficiency for B _j ⁿ that is a CU from all the encoding modes that can be used in the picture or a subset thereof. select.

The encoding control unit 1 generates, for example, a CU partition state as shown in FIG. 8 for each LCU used for encoding a picture to be encoded (current picture), and sets B _j ⁿ as a CU. Identify.
The shaded portion in FIG. 8A shows an area defined by B _j ⁿ , and FIG. 8B shows a situation where a coding mode m (B _j ⁿ ) is assigned by hierarchical division in a quadtree graph. Is shown.
In FIG. 8B, nodes surrounded by squares indicate nodes (CU) to which the encoding mode m (B _j ⁿ ) is assigned.

Note that B _j ⁿ that is a CU is further divided into PUs (Prediction Units) that are one or more partitions, as shown in FIG.
Hereinafter, the partition belonging to B _j ⁿ is ^denoted as P _ji ⁿ (i: PU number in the _j- ^th CU belonging to the nth layer).
How the partition P _ji ⁿ belonging to B _j ⁿ is divided is included as information in the encoding mode m (B _j ⁿ ).
All partitions P _ji ⁿ are subjected to prediction processing according to the encoding mode m (B _j ⁿ ), but individual prediction parameters can be selected for each partition P _ji ⁿ .

Changeover switch 3, the encoding control unit 1 selects the optimal coding mode m (B _j ⁿ⁾ for the partition P _ji ⁿ of each B _j ^n, the encoding mode m (B _j ⁿ⁾ is if intra-coding mode (step ST3), and outputs the partition P _ji ⁿ of B _j ⁿ a CU divided by the block division unit 2 to the intra prediction unit 4.
On the other hand, if the coding mode mB _j ⁿ is the inter coding mode (step ST3), the partition P _ji ⁿ of B _j ⁿ that is a CU divided by the block dividing unit 2 is output to the motion compensation prediction unit 5. .

Intra prediction unit 4 receives a partition P _ji ⁿ of B _j ⁿ from the changeover switch 3 is CU, intra prediction parameter corresponding to the encoding control unit 1 by the selected coding mode m (B _j ⁿ⁾ used, to implement the intra prediction process for partition P _ji ⁿ of B _j ^n, it generates an intra prediction image P _ji ⁿ (step ST4).
Intra prediction unit 4, when generating an intra prediction image P _ji ^n, the intra prediction image P _ji ⁿ outputs to the subtraction unit 6 and the addition unit 9, but the same intra prediction image P _ji in the video decoding apparatus of FIG. 4 ^{In order} to be able to generate ⁿ , the intra prediction parameter is output to the variable length coding unit 13.
The intra prediction process in the first embodiment is an AVC / H. When performing adaptive space prediction having directionality as defined in the H.264 standard (ISO / IEC 14496-10), information such as prediction mode information selected for each PU is included.
The intra prediction process in the first embodiment is an AVC / H. Although not limited to the algorithm defined in the H.264 standard, the intra-prediction parameter needs to include information necessary for generating exactly the same intra-predicted image on the moving image encoding device side and the moving image decoding device side. .

The motion compensation prediction unit 5 receives the partition P _ji ⁿ of B _j ⁿ from the changeover switch 3 is CU, from the B _j ⁿ of the partition P _ji ⁿ and the reference image stored in the motion compensated prediction frame memory 12 A motion vector is searched, using the inter prediction parameter corresponding to the motion vector and the encoding mode m (B _j ⁿ ), a motion compensated prediction process is performed on the partition P _ji ⁿ of the B _j ⁿ to perform inter prediction. It generates an image P _ji ⁿ (step ST5).

The motion compensation prediction unit 5 and generates an inter prediction image P _ji ^n, but outputs the inter prediction image P _ji ⁿ to the subtraction unit 6 and the addition unit 9, same inter prediction in a video decoding apparatus of FIG. 4 the image P _In ^order to be able to generate _jin , the inter prediction parameter is output to the variable length coding unit 13.
The inter prediction parameters used for generating the inter prediction image include the following information, and are multiplexed into the bitstream by the variable length encoding unit 13 in order to generate the exact same inter prediction image on the video decoding device side. Is done.
If, CU is in the B _j ⁿ partitioning a configuration including a plurality of reference images to the motion vector and motion compensation prediction frame memory 12 of the partition P _ji ⁿ is described mode information-PU, any reference picture Reference image indication index information that indicates whether to perform prediction using-Index information that indicates which motion vector prediction value to select and use when there are multiple motion vector prediction value candidates-Within multiple motion compensation Index information indicating which filter to select and use when there is an insertion filter. The motion vector of the partition P _ji ⁿ which is the PU has a plurality of pixel accuracy (half pixel, 1/4 pixel, 1/8 pixel). Selection information indicating which pixel precision to use.

When the intra prediction unit 4 or the motion compensation prediction unit 5 generates a prediction image (intra prediction image P _ji ⁿ , inter prediction image P _ji ⁿ ), the subtraction unit 6 is B _j that is a CU divided by the block division unit 2. ^A subtraction image is generated by subtracting the prediction image (intra prediction image P _ji ⁿ , inter prediction image P _ji ⁿ ) generated by the intra prediction unit 4 or the motion compensation prediction unit 5 from the partition P _ji ^{n of n} , It outputs the prediction difference signal e _ji ⁿ indicating the differential image to the transform and quantization unit 7 (step ST6).

Transform and quantization unit 7 receives the prediction difference signal e _ji ⁿ from the subtracting unit 6, the transform block size unit that is included in the predictive differential coding parameter output from the coding controller 1, a plurality of prediction The difference signal e _ji ⁿ is collected to obtain a prediction difference signal e _i ⁿ , and the prediction difference signal e _i ⁿ is converted based on the prediction difference encoding parameter (for example, DCT (discrete cosine transformation), A transform coefficient is calculated by performing an orthogonal transform process such as a KL transform in which a base design is made for a specific learning sequence.
Further, the transform / quantization unit 7 quantizes the transform coefficient using the quantization parameter included in the prediction difference encoding parameter, thereby converting the quantized transform coefficient into the compressed data of the difference image. Is output to the inverse quantization / inverse transform unit 8 and the variable length coding unit 13 (step ST7).

Here, as a unit for performing the orthogonal transformation, a TU (Transform Unit: orthogonal transformation processing unit block) unit obtained by dividing with a quadtree structure starting from the CU is applied.
The encoding control unit 1 also determines the maximum TU size, the minimum TU size, and the maximum TU partition layer number (the number of partition layers in the CU with a quadtree structure where the layer is the deepest) as in the case of the LCU.
As a method of determining the TU size, for example, a method of determining a size corresponding to the resolution of the input image for all the pictures can be considered.
A method of quantifying the difference in complexity of local motion of an input image as a parameter, determining a TU size to a small value for a picture with intense motion, and determining a TU size to a large value for a picture with little motion And so on.
As for the maximum number of TU division layers, for example, the more the input image moves, the deeper the layer number is set so that orthogonal transformation can be detected in finer units. A method of setting to suppress can be considered.
Note that the maximum number of TU partition layers may be configured to use a fixed value in sequence units.
The transform / quantization unit 7 divides each CU hierarchically into TUs up to the maximum number of TU partition layers defined above, and performs the orthogonal transform and quantization on each TU.

When the inverse quantization / inverse transform unit 8 receives the compressed data of the difference image from the transform / quantization unit 7, the inverse quantization / inverse transform unit 8 uses the quantization parameter included in the prediction difference encoding parameter output from the encoding control unit 1. Then, the compressed data of the difference image is inversely quantized, and the inverse quantization processing (for example, inverse DCT (Inverse Discrete Cosine Transform) is performed on the transform block size unit included in the prediction difference encoding parameter. ) And inverse transformation processing such as inverse KL transformation), the compressed data after the inverse transformation processing is converted into a local decoded prediction difference signal e _ji ⁿ hat ("^" Is expressed as a hat) to the adder 9 (step ST8).

Predictive differential coding parameters outputted from the coding controller 1, for each area of the CU, the quantization parameter used for encoding the prediction difference signal e _ji ⁿ therein, include information indicating the division state of the TU Yes.
The prediction differential encoding parameter is determined together when the encoding control unit 1 determines the encoding mode m (B _j ⁿ ).
One quantization parameter may be assigned in units of LCUs and used in common in units of divided CUs, or may be expressed as a difference value from the LCU quantization parameter for each CU. .
Incidentally, division of the TU, the CU without may be configured to determine the P _ji ⁿ is a PU split from CU units.

When the adder 9 receives the local decoded prediction difference signal e _ji ⁿ hat from the inverse quantization / inverse transform unit 8, the adder 9 performs the local decoded prediction difference signal e _ji ⁿ hat and the intra prediction unit 4 or the motion compensated prediction unit 5. By adding a prediction signal indicating the generated prediction image (intra prediction image P _ji ⁿ , inter prediction image P _ji ⁿ ), the local decoded partition image P _ji ⁿ hat or a local decoded encoded block image as a collection thereof is used. A local decoded image is generated (step ST9).
When generating the locally decoded image, the adding unit 9 stores the locally decoded image signal indicating the locally decoded image in the intra prediction memory 10 and outputs the locally decoded image signal to the loop filter unit 11. The locally decoded image signal becomes an image signal for subsequent intra prediction.

  The processes of steps ST3 to ST9 are repeatedly performed until the processes for all the hierarchically divided CUs are completed, and when the processes for all the CUs are completed, the process proceeds to the process of step ST13 (steps ST10 and ST11).

When the loop filter unit 11 receives the local decoded image signal from the adder unit 9, the loop filter unit 11 performs a predetermined filtering process on the local decoded image signal, thereby reducing the encoding distortion included in the local decoded image signal. The local decoded image indicated by the local decoded image signal after compensation for coding distortion is stored in the motion compensated prediction frame memory 12 as a reference image (step ST12).
The filtering process by the loop filter unit 11 may be performed in units of LCUs or individual CUs of the local decoded image signal output from the adding unit 9, or a local decoded image signal corresponding to one screen CU is output. You may do it for one screen later.

The variable length encoding unit 13 includes an encoding mode, a prediction differential encoding parameter, an LCU size, a maximum CU partition layer number, a TU maximum size, a TU minimum size, a maximum TU partition layer number output from the encoding control unit 1 and The division flag (the CU division state of the LCU), the compressed data output from the transform / quantization unit 7 and the TU division state of each CU, and the intra prediction parameter or motion compensation prediction unit 5 output from the intra prediction unit 4 Entropy-encode the output inter prediction parameter.
The variable length coding unit 13 is a coding mode that is a coding result of entropy coding, a prediction differential coding parameter, an LCU size, a maximum number of CU partition layers, a TU maximum size, a TU minimum size, a maximum TU partition layer number, The bit stream is generated by multiplexing the division flag (the CU division state of the LCU), the compressed data, the TU division state of each CU, and the intra prediction parameter / inter prediction parameter encoded data (step ST13).
FIG. 9 is an explanatory diagram showing an example of a bit stream generated by the variable length encoding unit 13, and this bit stream includes the maximum number of CU partition layers in the encoded data of the LCU. Is shown.

Hereinafter, the “CU division state entropy encoding process”, which is a feature of the first embodiment, will be described in detail.
In the first embodiment, starting from LCU (hierarchy n = 0), quadtree division of each CU is recursively repeated until the maximum number of CU division hierarchies is reached.
First, the variable length encoding unit 13 encodes the maximum number of CU division layers (step ST21 in FIG. 3).

For the code of the maximum number of CU partition layers, in the case of an LCU belonging to a picture (or a slice obtained by dividing a picture) that only performs intra-frame coding, for example, a binarized value is determined by referring to the table on the left side of FIG. In the case of an LCU belonging to a slice or picture for which interframe coding is performed, for example, it is determined by referring to the table on the right side of FIG.
In intra-frame coding, it is difficult to predict and the LCU is easily divided finely. Therefore, a shorter code is assigned as the maximum number of CU division layers is larger.
On the other hand, since prediction is relatively easy in inter-frame coding, a shorter code is assigned as the maximum number of CU partition layers is smaller, so that the CU partition state can be efficiently encoded.
Further, as shown in FIG. 10, the table is switched according to the upper limit value of the number of divided CU layers set for each sequence, picture, or slice, and only indexes up to the upper limit value of the number of divided CU layers can be obtained. By using a table that does not have, the code redundancy can be minimized, and the division state of the CU can be efficiently encoded.

The variable length coding unit 13 proceeds to the 0th layer CU coding process (step ST22).
When the processing target layer n is 0 and the maximum number of CU partition layers is 2 or more (step ST201), the variable length encoding unit 13 performs quadtree partitioning on the CUs belonging to the processing target layer n. Without encoding the division flag indicating whether or not there is, the CU encoding processing of the CU belonging to the (n + 1) th hierarchy obtained by dividing the CU into the quadtree is executed in the Z-scan order (step ST205).
If the processing target layer n is not the deepest layer corresponding to the maximum number of CU partition layers (step ST202), the variable length encoding unit 13 performs quadtree partitioning on each CU belonging to the processing target layer n. A division flag indicating whether or not the image is present is encoded (step ST203).
For example, when the CU is divided into quadtrees, the division flag is encoded as “1”, and when the CU is not divided into quadtrees, the division flag is encoded as “0”. The division flag of the CU is encoded in the order of Z scan starting from the LCU, for example. If the CU has been divided into quadtrees (step ST204), the CU encoding processing of the CU belonging to the (n + 1) th layer obtained by dividing the CU into quadtrees is executed in the Z-scan order (step ST205). When the CU is not divided into quadtrees, the CU encoding process is terminated. At this time, the encoded data belonging to the CU may be encoded, or the encoded data of each CU may be encoded after encoding all the CU division states.
When the processing target layer n is the deepest layer corresponding to the maximum number of CU partition layers, the CU belonging to the layer is not further divided into quadtrees, so the CU partition flag is not encoded. Then, the CU encoding process ends.

When the maximum number of CU partition layers is not “1”, as described above, it is obvious that the LCU performs quadtree partitioning at least once. Therefore, the partition flag for the LCU is not encoded. .
As described above, when the maximum number of CU partition layers is not “1”, the partition flag for the LCU is not encoded, so that the CU partition state can be efficiently encoded.

In addition, when encoding the CU partition flag in the Z-scan order, a CU that has the maximum number of CU partition layers is generated before the last CU that is not encoded in the Z-scan order is encoded. If not, it is obvious that the last CU that is not encoded in the Z-scan order is divided up to the layer corresponding to the maximum number of CU division layers, and in this case also, the CU division flag is not encoded. ing.
As described above, since the CU division flag is not encoded for a CU that is obvious to be divided, the CU division state can be efficiently encoded.

As these entropy coding, arithmetic coding may be applied, or Huffman coding may be applied.
Hereinafter, the encoding process of the CU division state will be described using a specific example of FIG.
In the example of FIG. 11, when the LCU size is 64 pixels × 64 pixel size and the upper limit of the number of divided layers of the CU is 4 as the LCU in the slice or picture for performing interframe coding, that is, the CU size is 64 pixels × It is assumed that 64 pixels, 32 pixels × 32 pixels, 16 pixels × 16 pixels, and 8 pixels × 8 pixels can be selected.
In the example of FIG. 11, the LCU to be processed is divided into quadrants into CUs of 32 pixels × 32 pixels, and the upper left, upper right, and lower right 32 pixels × 32 pixels of CUs are further 16 pixels × 16 pixels in size. The CU is divided into quadtrees.

In the conventional method, since the LCU (layer 0) is divided into quadtrees, “1” is encoded as a CU division flag for the LCU.
Next, in the first layer, the division flag of the CU is encoded in the Z scan order. That is, the CU division flag is encoded for each CU having a size of 32 pixels × 32 pixels in the upper left, upper right, lower left, and lower right.
In the example of FIG. 11, “1” is encoded as the division flag for the CU having the size of 32 pixels × 32 pixels in the upper left, upper right, and lower right, and “0” is used as the division flag for the CU having the size of 32 pixels × 32 pixels in the lower left. ".
Since no further quadtree division is performed for a 16 pixel × 16 pixel size CU, “0” is encoded as a CU division flag for each CU of 16 × 16 pixel size.
Thus, in the conventional method, there are one CU partition flag in the 0th layer, 4 CU partition flags in the 1st layer, and 12 CU partition flags in the 2nd layer, a total of 17 The CU partition flag is encoded, and the encoding process of the CU partition state is completed.

On the other hand, in the method of the first embodiment, since the LCU is divided up to a CU having a size of 16 pixels × 16 pixels belonging to the second hierarchy, the total of the 0th hierarchy, the first hierarchy, and the second hierarchy “ “110” (three binarized values) corresponding to the maximum number of CU partition layers of 3 ”is encoded (see the table on the right side of FIG. 10). In this case, since the maximum number of CU division layers of the LCU is not “1” and it is obvious that the LCU is divided once, the CU division flag for the LCU (0th layer) is not encoded.
Next, in the first layer, the division flag of the CU is encoded in the Z scan order. That is, the division flag of the CU is encoded for each CU having a size of 32 pixels × 32 pixels in the upper left, upper right, lower left, and lower right.
In the example of FIG. 11, “1” is encoded as the division flag for the CU having the size of 32 pixels × 32 pixels in the upper left, upper right, and lower right, and “0” is used as the division flag for the CU having the size of 32 pixels × 32 pixels in the lower left. ".
In the example of FIG. 11, since the maximum number of CU division layers of an LCU is “3” and it is clear that no further CU division is performed, no more CU division flags are encoded.
As described above, in the method according to the first embodiment, three binarized values indicating the maximum number of CU partition layers and four CU partition flags in the first layer are to be encoded. Seven binarized values are encoded, and the encoding process in the CU division state is completed.

In the encoding process in the CU partitioning state, when applying arithmetic coding, the encoding of the maximum number of CU partitioning layers is binarized using the table of FIG. 10, and then each binarized value (bin) is set. On the other hand, by performing adaptive arithmetic coding with an appropriate conditional occurrence probability, the amount of codes can be further reduced.
The division state of the CU is formed depending on the motion and texture of the screen, but the screen motion and texture are generally likely to be similar in the vicinity of the space.
Therefore, for example, the maximum CU partition layer number on the upper side of the LCU to be encoded is MaxCUDepth_Above, the maximum CU partition layer number on the left side is MaxCUDepth_Left, and the maximum CU partition layer number MaxCUDepth_Above, MaxCUDepth according to the encoding of MaxCUDepth_LCU The code length of the code representing the maximum number of CU division layers may be switched.
That is, according to the maximum CU partition layer number MaxCUDepth_Above and MaxCUDepth_Left above and to the left of the LCU to be encoded, a memory that holds the occurrence probability of the dominant symbol of the first bin of the maximum CU partition layer number of the LCU to be encoded You may make it switch.

In the case of an LCU included in a slice or picture for which intra-frame coding is performed, there is a high possibility that the maximum number of CU partition layers is large. Therefore, both the upper maximum CU partition layer number MaxCUDepth_Above and the left maximum CU partition layer number MaxCUDepth_Left When taking a large value (for example, when both are not “0”, or when one of them is the upper limit of the number of divided layers of a CU, or when the sum of both is larger than a predetermined threshold), the LCU to be encoded Apply the dominant symbol and occurrence probability of the first bin so that the maximum CU partition hierarchy number of the first CU has a high probability of taking a large value.
On the other hand, when both the maximum CU partition layer number MaxCUDepth_Above on the upper side and the maximum CU partition layer number MaxCUDepth_Left on the left side do not take a large value, there is a probability that the maximum CU partition layer number of the encoding target LCU takes a large value. Apply the first bin dominant symbol and probability of occurrence to be low.
In this way, by switching the memory holding the occurrence probability of the dominant symbol of the first bin of the maximum CU partition layer number of the encoding target LCU, it is possible to encode a symbol with a high probability of occurrence with a small amount of information. Therefore, the CU division state can be efficiently encoded.

In addition, in the case of an LCU included in a slice or picture to be subjected to inter-frame coding, there is a high possibility that the maximum number of CU partition layers will be small, so the upper maximum CU partition layer number MaxCUDepth_Above and the left maximum CU partition layer number MaxCUDepth_Left When both take a small value (for example, when both are “0”, or when both are not the upper limit of the number of divided layers of a CU, or when the sum of both is smaller than a predetermined threshold value) Apply the first bin dominant symbol and occurrence probability that the maximum number of CU partition layers of the LCU is high.
On the other hand, when both the upper maximum CU partition layer number MaxCUDepth_Above and the left maximum CU partition layer number MaxCUDepth_Left do not take a small value, the probability that the maximum CU partition layer number of the LCU to be encoded takes a small value is high. Apply the first bin dominant symbol and probability of occurrence to be low.
In this way, by switching the memory holding the occurrence probability of the dominant symbol of the first bin of the maximum CU partition layer number of the encoding target LCU, it is possible to encode a symbol with a high probability of occurrence with a small amount of information. Therefore, the CU division state can be efficiently encoded.

  When Huffman coding is applied in the encoding process in the CU partitioning state, as shown in FIG. 12, the upper maximum CU partition layer number MaxCUDepth_Above and the left maximum CU partition layer number MaxCUDepth_Left are small and large. In some cases, by switching the Huffman table, in arithmetic coding, a short code is assigned to a coded symbol whose probability of occurrence of a dominant symbol is high, and in the opposite case, a long code is assigned. Similar effects can be obtained.

  In the first embodiment, the maximum CU partition layer number is always encoded for each LCU, but the upper maximum CU partition layer number MaxCUDepth_Above and the left maximum CU partition layer number MaxCUDepth_Left satisfy a predetermined condition. Only when the maximum CU partition layer number of the encoding target LCU is variable-length encoded, and if the predetermined condition is not satisfied, the maximum CU partition layer number of the encoding target LCU is not variable-length encoded. The division flags of all the CUs output from the block division unit 2 may be configured to be variable length encoded.

That is, when both the upper maximum CU partition layer number MaxCUDepth_Above and the left maximum CU partition layer number MaxCUDepth_Left take a small value (for example, both are “0”, or both are the upper limit of the number of CU partition layers) Otherwise, or when the sum of both is smaller than a predetermined threshold value), the CU partition state is encoded by the above-described method of encoding the maximum number of CU partition layers of the LCU to be encoded.
On the other hand, when both the upper maximum CU partition layer number MaxCUDepth_Above and the left maximum CU partition layer number MaxCUDepth_Left do not take a small value, the maximum CU partition layer number of the LCU to be encoded is not variable-length encoded. In addition, the CU partitioning state is encoded by variable length encoding the partition flags of all the CUs output from the block partitioning unit 2.
With this configuration, when a large number of LCUs having the maximum number of maximum CU partition layers occur accidentally, the overhead of the maximum number of CU partition layers can be reduced, and the CU partition state can be efficiently performed. Can be encoded.

  As can be seen from the above, according to the moving picture coding apparatus of FIG. 1, the variable length coding unit 13 has the maximum CU partitioning that indicates the number of split layers in the deepest part of the quadtree structure in each LCU. Since the number of hierarchies is variable-length coded, and the division flag indicating whether or not a CU other than the CU belonging to the deepest hierarchy is quadtree-divided by the block dividing unit 2 is variable-length coded, There is an effect that the division state of the CU can be transmitted to the video decoding device side by signaling with a small number of bits.

Next, processing contents of the image decoding apparatus in FIG. 5 will be described.
When the variable length decoding unit 51 receives the bit stream output from the image encoding device in FIG. 1, the variable length decoding unit 51 performs variable length decoding processing on the bit stream (step ST31 in FIG. 5), and from one or more frames of pictures. The frame size (the number of horizontal pixels and the number of vertical lines) is decoded in units of sequence units or pictures.
The variable length decoding unit 51 determines the LCU size and the maximum number of CU partition layers in the same procedure as the encoding control unit 1 in FIG. 1 (step ST32).
For example, in the image encoding apparatus, when the LCU size is determined according to the resolution of the input image, the LCU size is determined based on the previously decoded frame size.
In addition, when the LCU size and the maximum number of CU division layers are multiplexed in the bit stream, the information decoded from the bit stream is referred to.
The variable length decoding unit 51 also performs variable length decoding of the CU partition flag from the encoded data, and decodes the CU partition state indicating the LCU quadtree structure from the maximum number of CU partition layers and the CU partition flag. .

Hereinafter, the entropy decoding process in the CU division state will be described in detail.
First, the variable length decoding unit 51 decodes the maximum number of CU division layers with reference to the table of FIG. 10 (step ST51 of FIG. 6).
For example, when the upper limit of the number of CU partition layers is 4, if the encoded data is a binary value of “110” by interframe coding, decoding is performed assuming that the maximum number of CU partition layers is “3”. Is done.
The variable length decoding unit 51 proceeds to the 0th layer CU decoding process (step ST52).
When the processing target layer n is 0 and the maximum number of CU partition layers is 2 or more (step ST501), the variable length decoding unit 51 determines whether the CU belonging to the processing target layer n is divided into quadtrees. Without decoding the split flag indicating whether or not, the CU decoding process of the CU belonging to the (n + 1) th layer obtained by dividing the CU into the quadtree is executed in the Z-scan order (step ST505).

If the processing target layer n is not the deepest layer corresponding to the maximum number of CU partition layers (step ST502), the variable length decoding unit 51 decodes the partition flag of each CU belonging to the process target layer n (step ST502). ST503).
Note that the division flag of the CU is decoded in the same order as encoding. For example, decoding is performed in the order of Z scan starting from LCU.
When the division flag of the CU indicates that the CU is to be divided (step ST504), the CU decoding process of the CU belonging to the (n + 1) th hierarchy obtained by dividing the CU into the quadtree is executed in the Z scan order (step ST505). If the CU is not divided into quadtrees, the CU decoding process is terminated. At this time, the encoded data belonging to the CU may be decoded, or the encoded data of each CU may be decoded after decoding all the CU partition states.
However, since the CU belonging to the deepest hierarchy corresponding to the maximum number of CU division hierarchies is not subjected to further quadtree division, the CU decoding process is terminated without decoding the CU division flag.

When the maximum number of CU partition layers is not “1”, as described above, it is obvious that the LCU is subjected to quadtree partitioning at least once. Therefore, the partition flag for the LCU is not decoded.
Further, when decoding the CU partition flag in the Z scan order, if there is no CU that has the maximum number of CU partition layers before decoding the last CU that has not been decoded in the Z scan order, Since it is obvious that the last CU that has not been decoded in the Z-scan order is divided up to the layer corresponding to the maximum number of CU division layers, the division flag of the CU is not decoded in this case as well.
As a result, the CU partitioning state can be entropy-decoded, but in the video encoding device, the maximum number of CU partition layers of the encoding target LCU is output from the block partitioning unit 2 without being variable-length encoded. When the CU partitioning state is encoded by variable length coding of all the CU partition flags, the CU partition state is variable length decoded and the CU partition state is entropy decoded from the partition flag. .

When the variable length decoding unit 51 decodes the CU division state as described above, the variable length decoding unit 51 specifies B _j ⁿ that is each CU divided hierarchically from the CU division state (step ST33).
Further, the variable length decoding unit 51 decodes a coding mode m (B _j ⁿ ) corresponding to each CU B _j ⁿ from the coded data, and is included in the coding mode m (B _j ⁿ ). Based on the information, the CU is further divided into one or a plurality of PUs, and prediction parameters (encoding mode, compressed data, prediction differential encoding parameters, intra prediction parameters / inter prediction parameters, etc.) assigned to each PU are allocated. Decode (step ST34).
Note that when the coding mode assigned to the CU is the intra coding mode, the intra prediction parameter is decoded for each of one or more PUs belonging to the CU. Also, when the coding mode assigned to the CU is the inter coding mode, the inter prediction parameter is decoded for each of one or more PUs belonging to the CU.
Intra prediction parameter decoding is performed in the same procedure as the image encoding device, based on the intra prediction parameters of neighboring decoded PUs, the prediction value of the intra prediction parameter of the PU to be decoded is calculated, and the prediction value is used. To decrypt.

When the coding mode m (B _j ⁿ ) of the partition P _ji ⁿ belonging to B _j ⁿ which is the CU from the variable length decoding unit 51 is the intra coding mode (step ST35), the changeover switch 52 changes the variable length. The intra prediction parameter output from the decoding unit 51 is output to the intra prediction unit 53.
On the other hand, when the coding mode m (B _j ⁿ ) of the partition P _ji ⁿ is the inter coding mode (step ST35), the inter prediction parameter output from the variable length decoding unit 51 is output to the motion compensation prediction unit 54. .

When the intra prediction unit 53 receives the intra prediction parameter from the changeover switch 52, the intra prediction unit 53 performs an intra prediction process on the partition P _ji ⁿ of B _j ⁿ that is the CU using the intra prediction parameter, and performs the intra prediction image P _ji. ⁿ is generated (step ST36).
Upon receiving the inter prediction parameter from the changeover switch 52, the motion compensation prediction unit 54 performs a motion compensation prediction process on the partition P _ji ⁿ of B _j ⁿ that is a CU using the inter prediction parameter, and performs the inter prediction image. generating a P _ji ⁿ (step ST37).

The inverse quantization / inverse transform unit 55 outputs compressed data for each TU that is a transform processing unit from the variable length decoding unit 51, and includes the quantum included in the prediction difference encoding parameter output from the variable length decoding unit 51. The compressed data in TU units is inversely quantized using the quantization parameter.
Further, the inverse quantization / inverse transform unit 55 performs inverse transform processing (for example, inverse DCT (Inverse Discrete Cosine Transform)) on the compressed data of inverse quantization in units of transform block size included in the prediction differential encoding parameter. In addition, by performing reverse conversion processing such as reverse KL conversion), the compressed data after the reverse conversion processing is output to the adder 56 as a decoded prediction difference signal (a signal indicating a difference image before compression) (step ST38). .

The quantization parameter included in the predictive differential encoding parameter is restored in units of CUs from the encoded data extracted from the bitstream, and the TU size information is divided into four parts starting from the CU as in the case of LCU division. Extraction is performed from the bitstream in a format of division information expressed by tree division, a format in which selectable TU sizes are expressed as index information, and the like.
Based on this TU size information, the inverse quantization / inverse transform unit 55 specifies the block size of the transform / quantization process and performs the process. Note that the TU size information may be configured to be determined in units of PUs that divide a CU, not CUs.

  When the addition unit 56 receives the decoded prediction difference signal from the inverse quantization / inverse conversion unit 55, the addition unit 56 adds the decoded prediction difference signal and the prediction signal indicating the prediction image generated by the intra prediction unit 53 or the motion compensated prediction unit 54. Thus, a decoded image (a collection of one or a plurality of decoded PU images included in the CU) is generated, and a decoded image signal indicating the decoded image is stored in the intra prediction memory 57, and the decoding is performed. The image signal is output to the loop filter unit 58 (step ST39). The decoded image becomes an image signal for subsequent intra prediction.

The processes in steps ST33 to ST39 are repeated until the processes for all the hierarchically divided CUs are completed (step ST40).
When the loop filter unit 58 receives the decoded image signal from the adder unit 56, the loop filter unit 58 performs the same filtering process as the loop filter unit 11 of the image encoding device, thereby compensating for the encoding distortion included in the decoded image signal. Then, the decoded image indicated by the decoded image signal after coding distortion compensation is stored in the motion compensated prediction frame memory 59 as a reference image (step ST41).
The filtering process by the loop filter unit 58 may be performed in units of LCUs or individual CUs of the decoded image signal output from the adding unit 56, or after the decoded image signal corresponding to the LCU for one screen is output. You may do it all on the screen.

With the above configuration, the moving picture decoding apparatus in FIG. 4 can efficiently decode a coded bit stream and reproduce a video signal.
In the first embodiment, an efficient coding method for a CU partition state and a method for suitably decoding the CU partition state have been described. However, efficient coding and a suitable decoding are performed in the same manner for a TU partition state. It can be performed.

  In the present invention, any constituent element of the embodiment can be modified or any constituent element of the embodiment can be omitted within the scope of the invention.

  1 encoding control unit, 2 block dividing unit (block dividing unit), 3 changeover switch (predicted image generating unit), 4 intra prediction unit (predicted image generating unit), 5 motion compensation predicting unit (predicted image generating unit), 6 Subtraction unit (image compression means), 7 transformation / quantization unit (image compression means), 8 inverse quantization / inverse transformation unit, 9 addition unit, 10 intra prediction memory, 11 loop filter unit, 12 motion compensated prediction frame memory , 13 Variable length encoding unit (variable length encoding unit), 51 Variable length decoding unit (variable length decoding unit), 52 Changeover switch (prediction image generation unit), 53 Intra prediction unit (prediction image generation unit), 54 Motion Compensated prediction unit (predicted image generating unit), 55 Inverse quantization / inverse transform unit (decoded image generating unit), 56 Adder unit (decoded image generating unit), 57 Intra prediction memory, 58 Loop filter unit, 59 motion-compensated prediction frame memory.

Claims (10)

An input image is divided into macroblocks of a predetermined size, and each macroblock is hierarchically divided into a quadtree structure, thereby enabling a coding processing unit block that is a smaller block of coding units than the macroblock. And a prediction image generation for generating a prediction image by performing a prediction process on the encoding processing unit block in an encoding mode corresponding to the encoding processing unit block output from the block dividing means An image compression unit that compresses a difference image between the encoding processing unit block output from the block dividing unit and the prediction image generated by the prediction image generation unit, and outputs compressed data of the difference image; The compressed data output from the image compression means and the encoding mode are subjected to variable length encoding, and the compressed data and And a variable length coding means for generating a bitstream encoded data of the encoding mode are multiplexed,
The variable-length encoding means variable-length encodes the maximum number of divided hierarchies indicating the number of division hierarchies in the deepest part of the quadtree structure in each macroblock, and encodes processing units belonging to the deepest hierarchy A moving picture coding apparatus characterized by variable length coding a division flag indicating whether or not an encoding processing unit block other than a block has been subjected to quadtree division by the block division means.   The variable length encoding means, when variable length encoding the maximum number of division layers related to each macroblock, according to the maximum number of division layers related to the macroblock existing around the macroblock to be encoded, 2. The moving picture encoding apparatus according to claim 1, wherein the code length of a code representing the maximum number of division layers related to the encoding target macroblock is switched.   The variable-length encoding means is the maximum for the macroblock to be encoded only when the maximum number of division layers related to the macroblock existing above and to the left of the macroblock to be encoded satisfies a predetermined condition. If the number of division hierarchies is variable-length coded and the predetermined condition is not satisfied, the division flag for all coding processing unit blocks output from the block division means without variable length coding the maximum number of division hierarchies. The moving picture encoding apparatus according to claim 1 or 2, wherein the variable length encoding is performed. The block dividing means divides each encoding processing unit block hierarchically in a quadtree structure, thereby outputting an orthogonal transformation processing unit block that is a block of an orthogonal transformation unit smaller than the coding processing unit block. ,
The variable-length encoding means variable-length encodes the maximum number of divided hierarchies indicating the number of divided hierarchies in the deepest part of the quadtree structure in each encoding processing unit block, and orthogonal transforms belonging to the deepest hierarchies 2. The moving picture coding apparatus according to claim 1, wherein a division flag indicating whether or not an orthogonal transform processing unit block other than the processing unit block has been subjected to quadtree division by the block division means is variable length coded. . Variable length decoding means for variable length decoding compressed data and coding modes related to each encoding processing unit block hierarchically divided from the encoded data multiplexed in the bitstream, and the variable length decoding means Predictive image generating means for generating a prediction image by performing prediction processing on the encoding processing unit block in the encoding mode related to the encoding processing unit block variable-length decoded by the variable length decoding means, and variable length decoding by the variable length decoding means Decoded image generating means for generating a differential image before compression from the compressed data relating to the decoded encoded block, and adding the difference image and the predicted image generated by the predicted image generating means to generate a decoded image; With
The variable-length decoding means performs variable-length decoding on the maximum number of division layers related to each macroblock from the encoded data, and codes other than encoding processing unit blocks belonging to the deepest layer corresponding to the maximum number of division layers Variable length decoding is performed on a division flag indicating whether or not a unit block is divided into quadtrees, and a quadtree structure of the macroblock is decoded from the maximum number of division layers and the division flag. Video decoding device.   When the variable length decoding means performs variable length decoding on the maximum number of divided layers related to each macroblock, the variable length decoding unit calculates the macroblock to be decoded from the maximum number of divided layers related to the macroblock existing around the macroblock to be decoded. The moving picture decoding apparatus according to claim 5, wherein a code length of a code representing the maximum number of divided hierarchies is determined.   The variable-length decoding means is related to the macroblock to be decoded from the encoded data only when the maximum number of division layers related to the macroblock existing above and to the left of the macroblock to be decoded satisfies a predetermined condition. If the maximum number of division layers is variable-length decoded and a predetermined condition is not satisfied, the division flag for all the encoding processing unit blocks is variable-length decoded from the encoded data, and the macro to be decoded from the division flag 7. The moving picture decoding apparatus according to claim 5, wherein the quadtree structure of the block is decoded.   The variable-length decoding means performs variable-length decoding on the maximum number of division layers related to each encoding processing unit block from the encoded data, and other than orthogonal transform processing unit blocks belonging to the deepest hierarchy corresponding to the maximum number of division layers Variable length decoding of a division flag indicating whether or not the orthogonal transform processing unit block is divided into quadtrees, and decoding the quadtree structure of the encoding processing unit block from the maximum number of division layers and the division flag. The moving picture decoding apparatus according to claim 5. The block dividing means divides the input image into macroblocks of a predetermined size and hierarchically divides each macroblock with a quadtree structure so that it is a block with a smaller encoding unit than the macroblock. A block division processing step for outputting an encoding processing unit block; and a prediction image generation means for predicting the encoding processing unit block in an encoding mode corresponding to the encoding processing unit block output in the block division processing step. A predicted image generation processing step that performs processing to generate a predicted image, and an image compression unit includes the encoding processing unit block output in the block division processing step and the predicted image generated in the predicted image generation processing step. An image compression processing step of compressing the difference image and outputting compressed data of the difference image; The variable length coding means performs variable length coding on the compressed data and the coding mode output in the image compression processing step, and the compressed data and the coded data of the coding mode are multiplexed. Variable length encoding processing step for generating
In the variable-length encoding processing step, the maximum number of division layers indicating the number of division layers in the deepest part of the quadtree structure in each macroblock is variable-length encoded and the encoding process belonging to the deepest layer A moving picture coding method characterized in that variable length coding is performed on a division flag indicating whether or not an encoding processing unit block other than a unit block has been subjected to quadtree division in the block division processing step. Variable length decoding processing step in which variable length decoding means performs variable length decoding of compressed data and encoding mode related to each encoding processing unit block hierarchically divided from encoded data multiplexed in a bitstream And the prediction image generating means performs prediction processing on the encoding processing unit block in the encoding mode related to the encoding processing unit block variable-length decoded in the variable length decoding processing step to generate a prediction image. A prediction image generation processing step, and a decoded image generation means generates a difference image before compression from the compressed data related to the encoded block variable-length decoded in the variable length decoding processing step, and generates the difference image and the prediction image A decoded image generation processing step of generating a decoded image by adding the predicted image generated in the processing step,
In the variable-length decoding processing step, the maximum number of division layers related to each macroblock is variable-length decoded from the encoded data, and other than encoding processing unit blocks belonging to the deepest layer corresponding to the maximum number of division layers A variable length decoding is performed on a division flag indicating whether or not an encoding processing unit block is divided into quadtrees, and a quadtree structure of the macroblock is decoded from the maximum number of division layers and the division flag. A moving picture decoding method.

Images