JP2011223315A - Image encoding apparatus, image decoding apparatus, and method for image encoding and decoding - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide an image encoding apparatus having an improved encoding efficiency.SOLUTION: An image reduction section 201 reduces a block image based on a reduction bock size outputted by a reduction block size selection section 205. A prediction image conversion quantization section 202 outputs a first in-screen prediction data from the reduced image. A prediction image inverse quantization inverse conversion section 203 outputs a local decoded prediction image from the first in-screen prediction data. A prediction image enlargement section 204 enlarges the local decoded prediction image and generates a first in-screen prediction image. A difference image generation section 102 outputs a difference image between the block image and the first in-screen prediction image. A difference conversion quantization section 103 performs frequency conversion and quantization of the difference image and outputs a conversion coefficient. An entropy encoding section 104 entropy-encodes the first in-screen prediction data and the conversion coefficient and outputs a bitstream.

Description

  The present invention relates to an image coding apparatus, an image coding method, an image decoding apparatus for decoding compressed data, and an image decoding method that perform intra prediction when an image is compressed and transmitted.

  For example, H.264 recommended by ITU-T (International Telecommunication Union Telecommunication Standardization Sector). In the H.264 moving image encoding method, intra-frame redundancy is removed by performing intra-screen prediction, and a prediction error signal is compression-encoded. Conventionally, there has been an apparatus as disclosed in Patent Document 1, for example, as such an encoding apparatus that performs compression encoding and a decoding apparatus that decodes the encoding apparatus.

  In such a conventional image encoding device, in order to generate an encoded bit stream from an input image, first, the input image is a macroblock that is an input image having a macroblock size that is a predetermined encoding unit by an image dividing unit. The image is divided into input images and processed in units of macroblocks. The macroblock input image is subtracted from the prediction image generated by the intra prediction unit by the subtraction unit, and a difference image signal obtained as a result is output. The difference image signal is input to the frequency conversion unit, converted to a conversion coefficient representing the amplitude for each frequency by performing frequency conversion, and output to the quantization unit. The transform coefficient is quantized in the quantization unit, and the quantized transform coefficient is output. The output quantized transform coefficient is entropy-encoded in the entropy encoding unit and output as an encoded bit stream. The quantized transform coefficient is also input to the inverse quantization unit. The inverse quantization unit inversely quantizes the quantized transform coefficient and outputs a decoded transform coefficient. The decoded transform coefficient is inversely frequency-converted by an inverse frequency converter, thereby obtaining a decoded difference image. The decoded difference image is added to the predicted image by the adding unit, and is stored in the frame memory as a locally decoded image. The local decoded image stored in the frame memory is used by the prediction unit to generate a predicted image.

H. performed in the prediction unit. In the H.264 intra-screen prediction method, a predicted image is generated by linearly expanding a locally decoded image of an encoded block adjacent to a block to be encoded. At this time, in the intra 4 × 4 prediction mode and the intra 8 × 8 prediction mode, the optimal prediction direction is selected from the eight types of directions and the DC prediction in which prediction is performed using the average value of surrounding pixels. select. In the intra 16 × 16 prediction mode, an optimal one is selected from a total of four types of horizontal direction, vertical direction, diagonal direction, and DC prediction.
The selected prediction mode and prediction direction are each entropy-encoded in the entropy encoder and multiplexed into the encoded bitstream.

International Publication No. 08/07500

  However, in the conventional image encoding device, as described above, H.264 is used. In the H.264 intra-frame prediction, a predicted image is generated by linearly expanding a local decoded image around a block to be encoded. Therefore, a predicted image in which a change in the peripheral local decoded image is formed as a straight edge as it is is generated. Generated. Therefore, when the edge of the straight edge-like predicted image does not match the edge of the actual input image, a straight edge is added to the prediction error signal, resulting in a decrease in coding efficiency. That is, the conventional intra-frame prediction has a problem that the prediction efficiency is good for a video having an edge in a specific direction, and the prediction efficiency is poor for a video having an edge in another direction.

  The present invention has been made to solve the above-described problems, and obtains an image coding apparatus capable of performing stable and efficient prediction regardless of the texture of the input image and improving the coding efficiency. For the purpose.

  An image encoding device according to the present invention receives image data, divides the image into a plurality of blocks, outputs a block image, and inputs a block image to generate a first intra prediction image and a screen A first intra prediction unit that outputs first intra prediction data, which is information for reconstructing the intra prediction image, and a difference calculation between the block image and the first intra prediction image and outputting the difference image A difference image generation unit that performs frequency conversion and quantization on the difference image and outputs a transform coefficient, and an entropy coding unit that entropy codes the first intra prediction data and the transform coefficient and outputs a bitstream The first intra prediction unit determines a block size of a reduced image obtained by reducing the block image and outputs the reduced block size as a reduced block size. A selection unit; an image reduction unit that reduces the block image based on the reduced block size and outputs the reduced image; and a predicted image conversion quantization unit that frequency-converts and quantizes the reduced image and outputs first in-screen prediction data The first intra prediction data by inverse quantization and inverse frequency transform and outputting the local decoded prediction image, and the local decoded prediction image is enlarged, and the first intra prediction And a predicted image enlargement unit that generates an image.

  The image coding apparatus of the present invention determines a block size of a reduced image obtained by reducing a block image and outputs the reduced block size as a reduced block size, and reduces and reduces the block image based on the reduced block size. An image reduction unit for outputting an image, a prediction image conversion quantization unit for frequency-converting and quantizing the reduced image and outputting first predicted intra-screen data, and an inverse quantization and inverse frequency transform for the first intra-screen prediction data A prediction image inverse quantization conversion unit that outputs a local decoded prediction image and a prediction image enlargement unit that expands the local decoded prediction image and generates a first in-screen prediction image. Therefore, stable and efficient prediction can be performed, and coding efficiency can be improved.

It is a block diagram which shows the structure of the image coding apparatus which concerns on Embodiment 1 of this invention. It is a flowchart which shows operation | movement of the image coding apparatus which concerns on Embodiment 1 of this invention. It is a flowchart which shows operation | movement of the 1st intra estimation part of the image coding apparatus which concerns on Embodiment 1 of this invention. It is a block diagram which shows the modification of the image coding apparatus which concerns on Embodiment 1 of this invention. It is explanatory drawing which shows the structure of the bit stream of the image coding apparatus which concerns on Embodiment 1 of this invention. It is a block diagram which shows the structure of the image decoding apparatus which concerns on Embodiment 2 of this invention. It is a flowchart which shows operation | movement of the image decoding apparatus which concerns on Embodiment 2 of this invention. It is a flowchart which shows operation | movement of the 1st intra estimated image generation part of the image decoding apparatus which concerns on Embodiment 2 of this invention. It is a block diagram which shows the modification of the image decoding apparatus which concerns on Embodiment 2 of this invention.

Embodiment 1 FIG.
FIG. 1 is a block diagram showing an image coding apparatus according to Embodiment 1 of the present invention.
FIG. 1 is a block diagram showing a configuration of an image encoding device 100 according to Embodiment 1. The image encoding device 100 includes a block division unit 101, a first intra prediction unit (first intra prediction unit) 200, a difference image generation unit 102, a difference transform quantization unit 103, and an entropy encoding unit. 104, a difference inverse transform inverse quantization unit 105, an addition unit 106, a loop filter unit 107, a local decoded image storage unit 108, a motion compensation prediction unit 109, and a prediction method determination unit 110.

  The block dividing unit 101 inputs a moving image as an original image, divides the moving image into blocks, and generates an image of each divided block (hereinafter referred to as a block image) as a difference image generating unit 102, a first image To the intra prediction unit 200 and the motion compensation prediction unit 109. At this time, the block size may be a fixed size of 16 pixels × 16 pixels, for example, or may be a size adaptively selected from 64 pixels × 64 pixels, 32 pixels × 32 pixels, and 16 pixels × 16 pixels. Also good. The first intra prediction unit 200 performs intra-frame prediction to generate a first intra predicted image and output the first intra predicted image to the prediction method determination unit 110, and also entropy encodes information about the predicted image as a predicted image transform coefficient. Output to the unit 104. The motion compensation prediction unit 109 generates an inter prediction image by performing inter-frame prediction by performing motion compensation prediction using the local decoded image stored in the local decoded image storage unit 108 and the input block image. The prediction method determining unit 110 outputs the motion vector and information specifying the referenced local decoded image to the entropy encoding unit 104 as motion information.

  The prediction method determination unit 110 compares the prediction performance or the coding performance based on the intra-frame prediction performed by the first intra prediction unit 200 and the inter-frame prediction performed by the motion compensation prediction unit 109, and determines and determines the optimal prediction method. Is output to the difference image generation unit 102 and the addition unit 106 as a block prediction image, and the selected prediction method is output to the entropy encoding unit as a block mode. The difference image generation unit 102 takes the difference between the block image from the block division unit 101 and the block prediction image from the prediction method determination unit 110 and outputs the difference to the difference transform quantization unit 103 as a difference block image. The difference transform quantization unit 103 performs orthogonal transform such as DCT on the transform block size unit specified by transform block size information input to the difference block image from the difference image generation unit 102 and the generated orthogonal transform The coefficient is quantized according to the quantization step size determined by the input quantization parameter, and the quantized coefficient is output to the entropy coding unit 104 and the inverse difference transform inverse quantization unit 105. The entropy encoding unit 104 entropy-encodes the predicted image transform coefficient, motion information, block mode, quantization coefficient, and quantization parameter by variable-length encoding or arithmetic encoding, and outputs the result as a bitstream.

  The inverse difference transform inverse quantization unit 105 inversely quantizes the quantization coefficient in accordance with the quantization step size determined by the quantization parameter to obtain an inverse quantization orthogonal transform coefficient, and generates the generated inverse quantization orthogonal transform coefficient. The inverse orthogonal transform is performed, and the locally decoded difference block image is output to the adding unit 106. The addition unit 106 outputs the local decoded block image to the loop filter unit 107 by adding the block prediction image and the local decoded difference block image. The loop filter unit 107 uses the local decoded block image and the local decoded image stored in the local decoded image storage unit 108 to perform, for example, an H.D. A filtering process such as a deblocking filter process used in H.264 is performed, and the result is output to the local decoded image storage unit 108 as a local decoded image. The local decoded image storage unit 108 stores one or more local decoded images and uses them to generate a predicted image.

  As shown in FIG. 1, the first intra prediction unit 200 includes an image reduction unit 201, a predicted image transform quantization unit 202, a predicted image inverse quantization inverse transform unit 203, a predicted image enlargement unit 204, and a reduced block size selection unit 205. , A reduced image quantization parameter output unit 206 and a transform block size output unit 207. The image reducing unit 201 reduces the input block image at a reduction rate based on the reduced block size information output from the reduced block size selecting unit 205, and outputs the reduced image to the predicted image transform quantization unit 202. Image reduction methods include, for example, a method in which an average value of pixel values for every 2 × 2 pixels, 4 × 4 pixels, or 8 × 8 pixels is used as a pixel value of a reduced image, or a method of sub-sampling after applying a low-pass filter Any method may be used as long as the input image is reduced to generate a low-resolution image. The predictive image transform quantization unit 202 performs orthogonal transform such as DCT on the reduced image input from the image reduction unit 201 and outputs the generated orthogonal transform coefficient from the reduced image quantization parameter output unit 206. The image is quantized based on the reduced image quantization parameter, and is output to the entropy coding unit 104 and the predicted image inverse quantization inverse transform unit 203 as a predicted image transform coefficient. The predictive image inverse quantization inverse transform unit 203 performs inverse quantization and inverse orthogonal transform, which are inverse processes of the predictive image transform quantization unit 202, on the predictive image transform coefficient input from the predictive image transform quantizer 202. As a result, a locally decoded reduced image is generated and output to the predicted image enlargement unit 204. In the predicted image enlarging unit 204, a process for improving the resolution is performed on the locally decoded reduced image input from the predicted image inverse quantization inverse transform unit 203, and the first image is enlarged so as to have the same size as the block image. The prediction method determination unit 110 outputs the intra prediction image. Here, for example, the processing content may be enlarged by interpolation processing, or the processing content is not particularly limited as long as it is a processing that improves the resolution of an input image by any other method.

The reduced block size selection unit 205 determines a reduced block size that specifies the block size after reduction in the image reduction unit 201, and outputs the reduced block size to the image reduction unit 201. The reduced block size may be any size as long as it is smaller than the block image, such as 1/2, 1/4, or 1/8 of the size of the input block image. For example, it is possible to determine a reduced block size by evaluating a prediction error and encoding performance by trying a plurality of reduced block sizes. In this way, an optimal reduction size can be selected and encoding can be performed efficiently. In this case, the reduced block size is also output to the entropy encoding unit 104 and configured to perform entropy encoding.
Alternatively, when the reduced image quantization parameter is large, it is considered that high frequency components are easily reduced by quantization, and it is considered that a prediction error does not easily occur even if the reduced image quantization parameter is large. However, when the reduced image quantization parameter is small, it may be determined to reduce the block to a larger block. In this way, an appropriate reduction size can be selected, and the amount of code can be reduced because there is no need to transmit the reduction size. Even if it is determined that the block is reduced to a smaller block when the quantization parameter is large and the block is reduced to a large block when the quantization parameter is small, the same effect can be obtained for the same reason.

In addition, because high-frequency components other than noise are locally low in high-resolution video, it can be assumed that prediction errors are unlikely to occur even if the block is reduced to a smaller block. You may comprise so that it may reduce to a block. Even in this case, an appropriate reduction size can be selected, and the amount of codes can be reduced because it is not necessary to transmit the reduction size.
Alternatively, an optimum reduced block size may be determined in slice units, picture units, or sequence units, and entropy coding may be performed using the reduced block size as header information. In this case, since the optimum reduced block size can be determined with a small code amount, efficient coding can be realized.

The reduced image quantization parameter output unit 206 determines a reduced image quantization parameter to be used when performing the quantization in the predicted image transform quantization unit 202 and the inverse quantization in the predicted image inverse quantization inverse transform unit 203 to perform the predicted image conversion. The data is output to the quantization unit 202. It has been experimentally confirmed that the optimum reduced image quantization parameter in the first intra prediction unit 200 is different depending on the value of the quantization parameter used when the difference block image is quantized by the difference transform quantization unit 103. ing. When the quantization parameter is a large value, the difference block image is coarsely quantized, so that the finally obtained decoded image is an image in which the low frequency component is dominant. Therefore, when the reduced image quantization parameter is BPP_QP and the quantization parameter is QP, BPP_QP can be set in conjunction with QP to obtain optimum encoding performance. At this time, for example, a table describing the relationship between the reduced image quantization parameter BPP_QP and the quantization parameter QP is provided in advance, and a value obtained from the quantization parameter QP with reference to this table is set as the reduced image quantization parameter BPP_QP. Alternatively, BPP_QP obtained by the following equation (1) may be used as the predicted quantization step size.

BPP_QP = 27 (QP ≦ 41)
QP-15 (QP> 41) (1)

  In the example given here, the numbers and formulas themselves are not limited to this, but by learning a combination of BPP_QP and QP that maximizes the encoding performance, and creating a table or formula based on it Similar effects can be obtained. Further, since the optimum BPP_QP for the QP differs for each block according to the characteristics of the input signal, as another means, the optimum BPP_QP is determined and output by entropy encoding the value of the BPP_QP in the same manner as the QP. It may be. For example, a difference value between BPP_QP values in a block to be encoded can be encoded from a BPP_QP value in an adjacent block.

  The transform block size output unit 207 determines a transform block size that is a unit of orthogonal transform / inverse transform in the difference transform quantizer 103 and the difference inverse transform inverse quantizer 105, and determines the difference transform quantizer 103 and the inverse difference transform inverse. The data is output to the quantization unit 105. When the image signal contains a lot of high frequency components, the difference block image is divided into, for example, 4 × 4 pixel units or 8 × 8 pixel units so that the image signal can be expressed more precisely. Since the signal can be appropriately expressed by performing orthogonal transformation, high-performance coding can be realized. Conversely, when there are few high-frequency components, the overhead in entropy coding of the transform block increases, so that high-performance coding can be realized by performing orthogonal transform in units of large blocks such as 16 × 16 pixels. Therefore, the transform block size output unit 207 can realize high-performance coding by determining and outputting the optimum transform block size by comparing the coding performance of each transform block size. .

  Alternatively, if the quantization parameter is a large value, it is difficult to produce a high frequency component because coarse quantization is performed, so that the encoding is performed with a large transform block size, and the quantization parameter is a small value. Conversely, the encoding may be performed with a small transform block size. In this case, encoding is performed without entropy encoding the transform block size by using one or a plurality of threshold values so that the optimum transform block size can be selected according to the quantization parameter. It is also possible. In addition, when a small value is selected as the reduced block size, it is considered that the block image to be encoded has few high-frequency components. Therefore, when the reduced block size is a small value, If a large value is used and, conversely, if the reduced block size is a large value, a high-performance encoding can be realized even if a small value is used as the transform block size.

Next, the operation of the image coding apparatus 100 according to the first embodiment will be described. FIG. 2 is a flowchart showing the processing of the image coding apparatus 100, and the numerical value of each step number corresponds to the code of the block that performs the processing.
First, when a moving image is input to the image coding apparatus 100, the moving image is divided into blocks by the block dividing unit 101, and each divided block image is output (step ST101). Next, each block image is subjected to intra-frame prediction in the first intra prediction unit 200, the first intra predicted image is output, and information about the predicted image is output as a predicted image transform coefficient (step ST200). ). When a block image is input to the motion compensation prediction unit 109, an inter prediction image is generated and output by performing motion compensation prediction using the local decoded image input from the local decoded image storage unit 108. The information specifying the motion vector and the referenced local decoded image is output as motion information (step ST109).

  When the first intra prediction image and the inter prediction image obtained in this way are input to the prediction method determination unit 110, the prediction method determination unit 110 performs intra-frame prediction by the first intra prediction unit 200 and a motion compensation prediction unit. The optimal prediction method is determined by comparing the prediction performance or the encoding performance by inter-frame prediction according to 109, the prediction image by the determined method is output as a block prediction image, and the selected prediction method is set as the block mode. Output (step ST110). The output block prediction image is subjected to a difference calculation with the block image from the block dividing unit 101 in the difference image generation unit 102, and as a result, a difference block image is output (step ST102). The difference block image is subjected to orthogonal transform such as DCT in the transform block size unit specified by the transform block size information input from the transform block size output unit 207 in the difference transform quantization unit 103 and the generated orthogonal transform. The coefficient is quantized according to the quantization step size determined by the input quantization parameter, and the quantized coefficient is output (step ST103).

  The predicted image transform coefficient, motion information, block mode, quantization coefficient, and quantization parameter generated by the above processing are subjected to entropy coding such as variable length coding or arithmetic coding by the entropy coding unit 104, It is output as a stream (step ST104).

  On the other hand, the quantization coefficient is input to the inverse difference transform inverse quantization unit 105, and is inversely quantized according to the quantization step size determined by the quantization parameter to obtain an inversely quantized orthogonal transform coefficient, which is generated. The inversely quantized orthogonal transform coefficients are inversely orthogonally transformed, and a local decoded difference block image is output (step ST105). The locally decoded difference block image obtained in this way is added to the block prediction image in the adding unit 106, thereby generating and outputting a locally decoded block image (step ST106). When the generated local decoded block image is input to the loop filter unit 107, the local decoded block image and the local decoded image using the local decoded image stored in the local decoded image storage unit 108, for example, H.264. A filtering process such as the deblocking filter process used in H.264 is performed, and the result is output as a locally decoded image (step ST107). The local decoded image obtained in this way is stored in the local decoded image storage unit 108 and used for generating a predicted image (step ST108).

FIG. 3 is a flowchart showing details of the operation of the first intra prediction unit 200. Also in this figure, the numerical value of each step number corresponds to the code of the block that performs the process.
First, the reduced block size selection unit 205 determines and outputs a reduced block size that specifies a block size after reduction in the image reduction unit 201 (step ST205). Further, the reduced image quantization parameter output unit 206 determines and outputs the reduced image quantization parameter used when performing the quantization in the prediction image transform quantization unit 202 and the inverse quantization in the prediction image inverse quantization inverse transform unit 203. (Step ST206). Further, transform block size output section 207 determines and outputs a transform block size as a unit of orthogonal transform / inverse transform in difference transform quantizer 103 and difference inverse transform inverse quantizer 105 (step ST207).

  When the block image input to the first intra prediction unit 200 is input to the image reduction unit 201, the block image is reduced at a reduction rate based on the reduced block size information output from the reduced block size selection unit 205, and the reduced image is It is output (step ST201). Subsequently, when the reduced image is input to the predicted image transform quantization unit 202, the input reduced image is subjected to orthogonal transform such as DCT, and the generated orthogonal transform coefficient is reduced to the reduced image quantization parameter output unit 206. It is quantized based on the reduced image quantization parameter that is output more and is output as a predicted image transform coefficient (step ST202). The predicted image transform coefficient is generated and output in the predicted image inverse quantization inverse transform unit 203 by performing inverse quantization and inverse orthogonal transform, which are the inverse processes of the predicted image transform quantization unit 202, to generate a locally decoded reduced image. (Step ST203). The locally decoded reduced image is input to the predicted image enlargement unit 204, subjected to processing for improving the resolution, enlarged so as to have the same size as the block image, and output as the first intra predicted image (step ST204).

  In addition, when the block prediction image from the motion compensation prediction part 109 is selected in the prediction method determination part 110, you may comprise so that it may encode by using the value preset as conversion block size, The transform block size may be encoded by entropy encoding.

  In the above description, the reduced image quantization parameter, the reduced block size, and the transformed block size can be selected. However, even if the preset values are used for these, the reduced image is encoded and enlarged. By doing so, the rough structure of the block image can be appropriately predicted by obtaining the block prediction image, so that high-performance encoding processing can be performed. Of course, the same effect can be obtained even if only one or more of the reduced image quantization parameter, the reduced block size, and the transformed block size are made variable. Further, by configuring so that motion compensation prediction is not performed, the same effect can be obtained even when the present method is applied as a coding process for a still image.

  For example, as shown in FIG. The image encoding device 100a is configured to include the second intra prediction unit 200a that performs intra prediction by the H.264 encoding method, and is configured to select an optimal prediction mode from the first intra prediction and the second intra prediction. If so, higher performance encoding can be realized. Here, in FIG. 4, the configuration other than the addition of the second intra prediction unit 200 a is the same as that in FIG. 1, and thus the description thereof is omitted.

Next, the bit stream encoded by the image encoding device 100 will be described.
The input image is encoded by the image encoding device 100 in FIG. 1 based on the above-described processing, and is output as a bit stream in a block in which a plurality of blocks are bundled and in a unit in which a plurality of blocks are bundled (hereinafter referred to as a slice). Is done. FIG. 5 shows the data arrangement of the bit stream. The bit stream is configured as a collection of encoded data for the number of blocks included in a frame, and the blocks are unitized in units of slices. A picture level header for referencing blocks belonging to the same frame as common parameters is prepared, and block size information is stored in the picture level header. If the block size is fixed in sequence units higher than the picture level, the block size information may be multiplexed in the sequence level header.

  Each slice starts from a slice header, and subsequently, encoded data of each block in the slice is arranged. In the example of FIG. 5, it is shown that K blocks are included in the second slice. The block data is composed of a block header and the quantization coefficient. The block header includes a block mode, motion information, a quantization parameter, a predicted image transform coefficient, a reduced image quantization parameter, a reduced block size, and a transformed block size. Are arranged. It should be noted that the effect of the present invention can be obtained even if the blocks are configured as hierarchically configured macroblocks.

  As described above, according to the image coding apparatus of the first embodiment, the image data is input, the block dividing unit that outputs the block image by dividing the image data, the block image is input, and the first screen is displayed. A first intra prediction unit that generates first intra prediction data, which is information for generating an intra prediction image and reconstructing the intra prediction image, and a block image and a first intra prediction image A difference image generation unit that performs a difference operation and outputs a difference image, a difference transform quantization unit that frequency-converts and quantizes the difference image and outputs a transform coefficient, an entropy-coded bit in the first screen prediction data and a transform coefficient, and a bit An entropy encoding unit that outputs a stream, and the first intra prediction unit determines the block size of the reduced image obtained by reducing the block image and outputs it as the reduced block size A reduced block size selection unit, an image reduction unit that reduces the block image based on the reduced block size and outputs a reduced image, and a predicted image that frequency-converts and quantizes the reduced image and outputs first in-screen prediction data A transform quantization unit, a prediction image inverse quantization inverse transform unit that performs inverse quantization and inverse frequency transform on the first intra-screen prediction data, and outputs a local decoded prediction image; The prediction image enlargement unit that generates the intra-screen prediction image is provided, so that stable and efficient prediction can be performed regardless of the texture of the input image, and the encoding efficiency can be improved.

  Further, according to the image coding apparatus of Embodiment 1, the entropy coding unit entropy codes the reduced block size in a predetermined unit and outputs it as a bit stream, so that an optimum reduced size is selected. Can be encoded efficiently.

  Further, according to the image coding apparatus of the first embodiment, the reduced block size selection unit sets a small value when the quantization step size used in the prediction image transform quantization unit is large, and the quantization step size is Outputs the reduced block size that is the opposite of the size relationship of the quantization step size so as to increase the value when the size is small, so that an appropriate reduced size can be selected and the reduced size is transmitted Since there is no need, the amount of codes can be reduced.

  Also, according to the image coding apparatus of the first embodiment, the reduced block size selection unit sets a small value when the quantization step size used by the difference transform quantization unit is large, and the quantization step size is small. In some cases, to output a large value, a reduced block size that is opposite to the size relationship of the quantization step size is output, so an appropriate reduced size can be selected and the reduced size is transmitted Since there is no need, the amount of codes can be reduced.

  In addition, according to the image coding method of the first embodiment, the image data is input, the block dividing step for dividing the block into a plurality of blocks and outputting the block image, the block image is input, and the first intra prediction image A first intra prediction step for generating first intra prediction data, which is information for reconstructing the intra prediction image, and a difference calculation between the block image and the first intra prediction image A difference image generation step for performing a difference image generation and performing a frequency conversion and quantization on the difference image and outputting a transform coefficient, and entropy encoding the first in-screen prediction data and the transform coefficient to output a bit stream An entropy encoding step, and the first intra prediction step determines a block size of a reduced image obtained by reducing the block image. Reduced block size selection step for outputting as a small block size, image reduction step for reducing the block image based on the reduced block size and outputting the reduced image, frequency conversion and quantization of the reduced image, and first in-screen prediction data A predictive image transform quantization step for outputting, a predictive image inverse quantization inverse transform step for inversely quantizing and inverse frequency transforming the first intra prediction data, and outputting a local decoded predictive image, and a local decoded predictive image And a prediction image enlargement step for generating a first intra-screen prediction image, so that stable and efficient prediction can be performed regardless of the texture of the input image, and the encoding efficiency can be improved.

Embodiment 2. FIG.
FIG. 6 is a block diagram showing a configuration of image decoding apparatus 300 according to Embodiment 2. The image decoding apparatus 300 is an apparatus that suitably decodes the bitstream generated by the image encoding apparatus 100 according to Embodiment 1, and outputs a decoded image obtained as a result. The image decoding apparatus 300 and the first intra Prediction image generation unit (first in-screen prediction image generation unit) 400, difference inverse transform inverse quantization unit 305, addition unit 306, loop filter unit 307, local decoded image storage unit 308, motion compensation unit 309, a prediction method determination unit 310, and a prediction changeover switch 311. Here, the entropy decoding unit 304 to the prediction method determination unit 310 and the first intra prediction image generation unit 400 are respectively the entropy encoding unit 104 to the prediction method determination unit 110 and the first intra prediction image generation unit 400 in the image encoding device 100 illustrated in FIG. This is a configuration of a decoding device corresponding to one intra prediction unit 200.

  The entropy decoding unit 304 performs entropy decoding on the input bit stream by variable length decoding, arithmetic decoding, or the like, and each of the first intra predicted image as a predicted image transform coefficient, motion information, block mode, quantization coefficient, and quantization parameter The data is output to the generation unit 400, the motion compensation unit 309, the prediction method determination unit 310, and the inverse difference transform inverse quantization unit 305. The difference inverse transform inverse quantization unit 305 inversely quantizes the quantization coefficient according to the quantization step size determined by the quantization parameter to obtain an inverse quantization orthogonal transform coefficient, and generates the generated inverse quantization orthogonal transform coefficient. The inverse orthogonal transform is performed, and the locally decoded difference block image is output to the adding unit 306. The first intra predicted image generation unit 400 generates a first intra predicted image by performing intra prediction based on the predicted image conversion coefficient, and outputs the first intra predicted image to the adding unit 306. The motion compensation unit 309 generates an inter prediction image by performing motion compensation using the local decoded image stored in the local decoded image storage unit 308 and the input motion information, and outputs the inter predicted image to the addition unit 306. The prediction method determination unit 310 determines whether to output a first intra prediction image by the first intra prediction image generation unit 400 or an inter prediction image by the motion compensation unit 309 according to the block mode, and performs prediction switching. Switch 311 is switched.

  When the prediction method determination unit 310 determines that the block mode is a mode for performing the first intra prediction, the prediction changeover switch 311 outputs the output from the entropy decoding unit 304 to the first intra prediction image generation unit 400. And the prediction image generated by the first intra prediction image generation unit 400 is connected to be output to the addition unit 306 as a block prediction image, and conversely, the block mode is determined to be a mode for performing motion compensation. In such a case, the output from the entropy decoding unit 304 is connected to the motion compensation unit 309, and the prediction image generated by the motion compensation unit 309 is connected to be output to the addition unit 306 as a block prediction image. The addition unit 306 adds the block prediction image and the local decoded difference block image to output the local decoded block image to the loop filter unit. The loop filter unit 307 performs filtering processing on the local decoded block image and the local decoded image using the local decoded block image and the local decoded image stored in the local decoded image storage unit 108, and uses the result as the local decoded image. Output to the external and local decoded image storage unit 308. The local decoded image storage unit 308 stores one or more local decoded images and uses them to generate a predicted image.

  As shown in FIG. 6, the first intra predicted image generation unit 400 includes a predicted image inverse quantization inverse transform unit 403, a predicted image enlargement unit 404, a reduced block size selection unit 405, a reduced image quantization parameter output unit 406, a transform. The block size output unit 407 is configured. These predicted image inverse quantization inverse transform unit 403 to transform block size output unit 407 are decoding corresponding to the predicted image inverse quantization inverse transform unit 203 to transform block size output unit 207 in the image coding apparatus 100 shown in FIG. It is the structure of an apparatus.

  The predicted image inverse quantization inverse transform unit 403 performs inverse quantization and prediction on the basis of the reduced image quantization parameter input from the reduced image quantization parameter output unit 406 with respect to the predicted image transform coefficient input from the entropy decoding unit 304. By performing inverse orthogonal transformation, a locally decoded reduced image is generated and output to the predicted image enlargement unit 404. The predicted image enlargement unit 404 performs processing for improving the resolution so that the input local decoded reduced image has the same size as the block image from the reduced block size input from the reduced block size selection unit 405. Is output to the adding unit 306 as an intra predicted image. The reduced block size selection unit 405 outputs a reduced block size that specifies the block size before enlargement in the predicted image enlargement unit 404 to the predicted image enlargement unit 404.

  Here, the reduced block size operates so as to reproduce the value determined by the image coding apparatus 100 according to the first embodiment. For example, in the image coding apparatus 100 according to the first embodiment, when a reduced block size is determined by evaluating a prediction error and coding performance by trying a plurality of reduced block sizes, entropy coding is performed for each block. Since the value obtained by entropy decoding the reduced block size is directly output from the entropy decoding unit 304 to the predicted image enlargement unit 404, the reduced block size selection unit 405 is unnecessary. Alternatively, in a flat area, since the high-frequency component of the image is small, it is considered that a prediction error is unlikely to occur even if the image is reduced to a small block. When the image coding apparatus 100 is configured to determine to reduce to a larger block when the conversion parameter is small, a reduced block size obtained by the same processing is output. Alternatively, when the image coding apparatus 100 is configured to determine to reduce to a small block when the quantization parameter is large and to reduce to a large block when the quantization parameter is small, the reduced block size obtained by the same processing is used. Is output.

  In addition, since it can be assumed that high-frequency components other than noise are locally small in a high-resolution video, the image coding apparatus 100 is configured to reduce to a small block and conversely to a large block in a low-resolution video. , The reduced block size obtained by the same processing is output. Alternatively, when the image encoding apparatus 100 is configured to determine an optimal reduced block size in slice units, picture units, or sequence units, and to perform entropy encoding using the reduced block size as header information, the header information is entropy decoded. Since the value obtained in this way is directly output from the entropy decoding unit 304 to the predicted image enlarging unit 404, the reduced block size selecting unit 405 is unnecessary.

  The reduced image quantization parameter output unit 406 outputs the reduced image quantization parameter used when the inverse quantization in the prediction image inverse quantization inverse transformation unit 403 is performed to the prediction image inverse quantization inverse transformation unit 403. Here, the reduced image quantization parameter operates so as to reproduce the value determined by the image coding apparatus 100 according to the first embodiment. In the image coding apparatus 100, when the reduced image quantization parameter is set to BPP_QP and the quantization parameter is set to QP, when the BPP_QP is set to be linked to the QP by a table or formula, the same table or formula is used. By using this, BPP_QP is determined and output as a reduced image quantization parameter. Further, in the image encoding device 100, when the BPP_QP value is configured to be entropy encoded by the same method as QP, the value obtained by entropy decoding is predicted from the entropy decoding unit 304 by the predicted image inverse quantization inverse transform. Since the data is directly output to the unit 403, the reduced image quantization parameter output unit 406 is not necessary.

  The transform block size output unit 407 outputs a transform block size that is a unit of inverse orthogonal transform in the difference inverse transform inverse quantization unit 305 to the difference inverse transform inverse quantization unit 305. Here, the transform block size operates so as to reproduce the value determined by the image coding apparatus 100 according to the first embodiment. When the image coding apparatus 100 is configured to determine and output an optimum transform block size by comparing the coding performance of each transform block size, the value obtained by entropy decoding is the entropy decoding unit 304. Since the difference is directly output to the inverse inverse transform inverse quantization unit 405, the transform block size output unit 407 is unnecessary.

  Also, if the quantization parameter is a large value, it is difficult to produce high-frequency components because coarse quantization is performed, so it is configured to encode with a large transform block size, and when the quantization parameter is a small value On the contrary, in order to encode with a small transform block size, one or a plurality of thresholds are used, and the optimum transform block size can be selected from a plurality of transform block sizes according to the quantization parameter. When the image encoding device 100 is configured to perform encoding without encoding, a transform block size obtained by similar processing is output.

  In addition, when a small value is selected as the reduced block size, it is considered that the block image to be encoded has few high-frequency components. Therefore, when the reduced block size is a small value, When the image coding apparatus 100 is configured to use a large value and, on the contrary, to use a small value as the transform block size when the reduced block size is a large value, the transform block size obtained by the same processing is output. Is done.

Next, the operation of the image decoding apparatus 300 according to Embodiment 2 will be described. FIG. 7 is a flowchart showing the processing of the image decoding apparatus 300, and the numerical value of each step number corresponds to the code of the functional unit that performs the processing.
First, when a bitstream is input to the image decoding apparatus 300, the entropy decoding unit 304 performs entropy decoding such as variable length decoding or arithmetic decoding, and the predicted image transform coefficient, motion information, block mode, quantization, and the like. Coefficients and quantization parameters are output (step ST304). In the inverse difference transform inverse quantization unit 305, the quantization coefficient is inversely quantized according to the quantization step size determined by the quantization parameter to obtain an inverse quantization orthogonal transform coefficient, and the generated inverse quantization orthogonal transform coefficient is Inverse orthogonal transform is performed and a local decoded difference block image is output (step ST305).

  In addition, the prediction method determination unit 310 determines which of the first intra prediction image by the first intra prediction image generation unit 400 or the inter prediction image by the motion compensation unit 309 is output according to the block mode, and the prediction The changeover switch 311 is changed over (step ST310). When the block mode is the first intra predicted image, the first intra predicted image generation unit 400 generates and outputs the first intra predicted image by performing intra prediction based on the predicted image conversion coefficient. (Step ST400). On the other hand, when the block mode is an inter prediction image, the motion compensation unit 309 performs motion compensation using the local decoded image stored in the local decoded image storage unit 308 and the input motion information, thereby performing the inter prediction image. Is generated and output (step ST309).

  When the block prediction image obtained in this way is input to addition section 306, the local decoded block image is output by adding the local decoded difference block image (step ST306). In the loop filter unit 307, filtering processing is performed on the local decoded block image and the local decoded image using the local decoded block image and the local decoded image stored in the local decoded image storage unit 108, and the result is the local decoded image. (Step ST307). In the local decoded image storage unit 308, one or a plurality of local decoded images are stored and used for prediction image generation (step ST308).

FIG. 8 is a flowchart showing details of the operation of the first intra predicted image generation unit 400. Also in this figure, the numerical value of each step number corresponds to the code of the functional unit that performs the process.
The reduced block size selection unit 405 outputs a reduced block size that specifies the block size before enlargement in the predicted image enlargement unit 404 (step ST405). Further, reduced image quantization parameter output section 406 outputs reduced image quantization parameters used when performing inverse quantization in predicted image inverse quantization inverse transform section 403 (step ST406). Further, transform block size output section 407 outputs a transform block size that is a unit of inverse orthogonal transform in difference inverse transform inverse quantization section 305 (step ST407). The predicted image inverse quantization inverse transform unit 403 performs inverse quantization on the predicted image transform coefficient input from the entropy decoding unit 304 based on the reduced image quantization parameter input from the reduced image quantization parameter output unit 406. By performing inverse orthogonal transform, a locally decoded reduced image is generated and output (step ST403). The predicted image enlarging unit 404 performs a process of improving the resolution so that the input local decoded reduced image has the same size as the block image from the reduced block size input from the reduced block size selecting unit 405. Are output as intra prediction images (step ST404).

  As an image encoding apparatus, as shown in FIG. In the case where the second intra prediction unit 200a that performs intra prediction by the H.264 encoding method is provided, as shown in FIG. It is preferable that a second intra-predicted image generation unit 400a for generating an intra-predicted image corresponding to the H.264 encoding method is provided so that the prediction method determination unit 310 can also select the second intra prediction according to the block mode. Can be decrypted. In the image decoding apparatus 300a shown in FIG. 9, the configuration other than the addition of the second intra predicted image generation unit 400a is the same as that of the image decoding apparatus 300 shown in FIG. Description is omitted.

  As described above, according to the image decoding apparatus of the second embodiment, the entropy decoding unit that entropy-decodes the bitstream and outputs the transform coefficient and the first in-screen prediction data, and the transform coefficient is inversely quantized and inversed. A differential inverse transform inverse quantization unit that performs frequency conversion and outputs a differential decoded image, and first intra prediction image generation that reconstructs and outputs the first intra prediction image based on the first intra prediction data A difference decoded image adding unit that adds the difference decoded image and the first intra prediction image and outputs a decoded image, and the first intra prediction image generation unit includes the first intra prediction image A predicted image inverse quantization inverse transform unit that performs inverse quantization and inverse frequency transform on the data and outputs a decoded predicted image, a reduced block size selection unit that outputs a reduced block size, and a decoded predicted image based on the reduced block size Enlarge Since a prediction image enlargement unit for generating a first intra prediction image, it is possible to suitably decode the coded video bit stream generated by the image coding apparatus according to the first embodiment.

  In addition, according to the image decoding apparatus of the second embodiment, an entropy decoding unit that entropy-decodes a bitstream and outputs a transform coefficient, a reduced block size, and first in-screen prediction data; A differential inverse transform inverse quantization unit that performs inverse frequency transform and outputs a differential decoded image, and a first intra prediction image that reconstructs and outputs the first intra prediction image based on the first intra prediction data A generating unit, and a differential decoded image adding unit that adds the differential decoded image and the first intra prediction image and outputs a decoded image. Predictive image inverse quantization and inverse transform unit that performs inverse quantization and inverse frequency transform on prediction data and outputs decoded prediction image, and expands decoded prediction image based on reduced block size to generate first in-screen prediction image Predicted image enlargement Since with the door, it is possible to decode efficiently coded video bit stream, it is not necessary to generate a reduced block size as a decoding device, it is possible to simplify the configuration.

  Also, according to the image decoding apparatus of the second embodiment, the reduced block size selection unit sets the quantization step size to a small value when the quantization step size used in the prediction image inverse quantization inverse transform unit is large, and the quantization step size. Since the reduced block size having a magnitude relation opposite to the magnitude relation of the quantization step size is output so that the value is large when the value is small, the reduced block size can be suitably selected.

  Also, according to the image decoding apparatus of the second embodiment, the reduced block size selection unit sets a small value when the quantization step size used in the differential inverse transform inverse quantization unit is large, and the quantization step size is Since the reduced block size having a magnitude relation opposite to the magnitude relation of the quantization step size is output so as to output a large value when the value is small, the reduced block size can be suitably selected.

  In addition, according to the image decoding method of the second embodiment, an entropy decoding step for entropy decoding a bitstream and outputting transform coefficients and first intra prediction data, and inverse quantization and inverse frequency transform of the transform coefficients are performed. A differential inverse transform inverse quantization step for outputting a differential decoded image; a first intra prediction image generation step for reconstructing and outputting the first intra prediction image based on the first intra prediction data; A difference decoded image addition step of adding the difference decoded image and the first intra prediction image and outputting the decoded image, wherein the first intra prediction image generation step includes the first intra prediction data A predictive image inverse quantization inverse transform step that performs inverse quantization and inverse frequency transform and outputs a decoded predicted image, a reduced block size selection step that outputs a reduced block size, and a reduced block And a prediction image enlargement step for enlarging the decoded prediction image based on the noise and generating a first intra-screen prediction image. Therefore, the encoded image bitstream generated by the image encoding device according to the first embodiment is Decoding can be suitably performed.

  100, 100a Image coding apparatus, 101 Block division unit, 102 Difference image generation unit, 103 Difference transform quantization unit, 104 Entropy coding unit, 105 Difference inverse transform inverse quantization unit, 106 Addition unit, 107 Loop filter unit, 108 local decoded image storage unit 109 motion compensation prediction unit 110 prediction method determination unit 200 first intra prediction unit 200a second intra prediction unit 201 image reduction unit 202 prediction image transform quantization unit 203 prediction Image inverse quantization inverse transform unit, 204 Predictive image enlargement unit, 205 Reduced block size selection unit, 206 Reduced image quantization parameter output unit, 207 Transform block size output unit, 300, 300a Image decoding device, 304 Entropy decoding unit, 305 Difference inverse transform inverse quantization unit, 306 addition unit, 307 loop Filter unit, 308 Local decoded image storage unit, 309 Motion compensation unit, 310 Prediction method determination unit, 311 Prediction changeover switch, 400 First intra prediction image generation unit, 400a Second intra prediction image generation unit, 403 Reverse prediction image Quantization inverse transform unit, 404 predicted image enlargement unit, 405 reduced block size selection unit, 406 reduced image quantization parameter output unit, 407 transformed block size output unit.

Claims (10)

A block dividing unit for inputting image data, dividing the block into a plurality of blocks, and outputting a block image;
A first intra prediction unit that inputs the block image, generates a first intra prediction image and outputs first intra prediction data that is information for reconstructing the intra prediction image;
A difference image generation unit that performs a difference calculation between the block image and the first intra prediction image and outputs a difference image;
A difference transform quantizing unit for frequency transforming and quantizing the difference image and outputting transform coefficients;
An entropy encoding unit that entropy-encodes the first intra-screen prediction data and the transform coefficient and outputs a bitstream;
The first in-screen prediction unit
A reduced block size selection unit that determines a block size of a reduced image obtained by reducing the block image and outputs the reduced block size;
An image reduction unit that reduces the block image based on the reduced block size and outputs a reduced image;
A predicted image transform quantization unit that frequency-converts and quantizes the reduced image and outputs first in-screen prediction data;
A predictive image inverse quantization inverse transform unit that performs inverse quantization and inverse frequency transform on the first intra-screen prediction data and outputs a locally decoded prediction image;
An image encoding apparatus comprising: a predicted image enlarging unit that expands the locally decoded predicted image and generates a first intra-screen predicted image.   The image encoding apparatus according to claim 1, wherein the entropy encoding unit entropy encodes the reduced block size in a predetermined unit and outputs the reduced block size as a bit stream.   The reduced block size selection unit sets the quantization step size to a small value when the quantization step size used by the predictive image transform quantization unit is large, and to a large value when the quantization step size is small. 2. The image encoding apparatus according to claim 1, wherein a reduced block size having a magnitude relation opposite to the magnitude relation of the step size is output.   The reduced block size selection unit sets a small value when the quantization step size used in the differential transform quantization unit is large, and outputs a large value when the quantization step size is small. 2. The image encoding apparatus according to claim 1, wherein a reduced block size having a magnitude relation opposite to the magnitude relation of the step size is output. An entropy decoding unit that entropy decodes the bitstream and outputs transform coefficients and first intra prediction data;
A differential inverse transform inverse quantization unit that performs inverse quantization and inverse frequency transform on the transform coefficient and outputs a differential decoded image;
A first intra-screen prediction image generating unit that reconstructs and outputs a first intra-screen prediction image based on the first intra-screen prediction data;
A difference decoded image adding unit that adds the difference decoded image and the first intra prediction image and outputs a decoded image;
The first in-screen predicted image generation unit
A predicted image inverse quantization inverse transform unit for inversely quantizing and inverse frequency transforming the first intra-screen prediction data and outputting a decoded predicted image;
A reduced block size selection unit for outputting a reduced block size;
An image decoding apparatus comprising: a predicted image enlarging unit that expands the decoded predicted image based on the reduced block size and generates the first intra-screen predicted image. An entropy decoding unit that entropy decodes the bitstream and outputs a transform coefficient, a reduced block size, and first intra prediction data;
A differential inverse transform inverse quantization unit that performs inverse quantization and inverse frequency transform on the transform coefficient and outputs a differential decoded image;
A first intra-screen prediction image generating unit that reconstructs and outputs a first intra-screen prediction image based on the first intra-screen prediction data;
A difference decoded image adding unit that adds the difference decoded image and the first intra prediction image and outputs a decoded image;
The first in-screen predicted image generation unit
A predicted image inverse quantization inverse transform unit for inversely quantizing and inverse frequency transforming the first intra-screen prediction data and outputting a decoded predicted image;
An image decoding apparatus comprising: a predicted image enlarging unit that expands the decoded predicted image based on the reduced block size and generates the first intra-screen predicted image.   The reduced block size selection unit is configured to reduce the quantization step size when the quantization step size used in the prediction image inverse quantization inverse conversion unit is large and to increase the quantization step size when the quantization step size is small. 6. The image decoding apparatus according to claim 5, wherein a reduced block size having a magnitude relation opposite to the magnitude relation of the conversion step size is output.   The reduced block size selection unit is configured to output a small value when the quantization step size used in the differential inverse transform inverse quantization unit is large, and to output a large value when the quantization step size is small. 6. The image decoding apparatus according to claim 5, wherein a reduced block size having a magnitude relation opposite to the magnitude relation of the conversion step size is output. A block division step for inputting image data, dividing the block into a plurality of blocks, and outputting a block image;
A first intra-screen prediction step for inputting the block image, generating a first intra-screen prediction image and outputting first intra-screen prediction data which is information for reconstructing the intra-screen prediction image;
A difference image generation step of performing a difference calculation between the block image and the first intra prediction image and outputting a difference image;
A difference transform quantization step for frequency transforming and quantizing the difference image and outputting transform coefficients;
An entropy encoding step of entropy encoding the first intra prediction data and the transform coefficient and outputting a bitstream;
The first in-screen prediction step includes
A reduced block size selection step of determining a block size of a reduced image obtained by reducing the block image and outputting the reduced block size;
An image reduction step of reducing the block image based on the reduced block size and outputting a reduced image;
Predictive image transform quantization step of frequency-converting and quantizing the reduced image and outputting first in-screen prediction data;
A predictive image inverse quantization inverse transform step of inversely quantizing and inverse frequency transforming the first intra-screen prediction data and outputting a locally decoded predicted image;
An image encoding method comprising: a predicted image enlargement step of enlarging the local decoded predicted image to generate a first intra-screen predicted image. An entropy decoding step of entropy decoding the bitstream and outputting transform coefficients and first in-screen prediction data;
Inverse quantization and inverse frequency transform of the transform coefficient, and a differential inverse transform inverse quantization step for outputting a differential decoded image;
A first intra prediction image generation step for reconstructing and outputting a first intra prediction image based on the first intra prediction data;
A difference decoded image addition step of adding the difference decoded image and the first intra prediction image and outputting a decoded image;
The first intra prediction image generation step includes
A predictive image inverse quantization inverse transform step of inversely quantizing and inverse frequency transforming the first in-screen prediction data and outputting a decoded predicted image;
A reduced block size selection step for outputting a reduced block size;
An image decoding method comprising: a predicted image enlargement step of enlarging the decoded predicted image based on the reduced block size and generating the first intra-screen predicted image.

Images