Image encoding device and method

FIELD: information technology.

SUBSTANCE: device for encoding moving images for increasing the compression ratio of a moving image signal while maintaining high image quality. To minimise the number of bits required for a residual unit, quantised conversion coefficients are adaptively divided into a plurality of subsets in accordance with the size of the conversion unit, and non-zero quantised conversion coefficients of each subset are scanned and encoded. The intra-prediction mode of the current unit is also determined using reference pixels obtained by filtering reference pixels of the current unit. Therefore, minimisation of the amount of data required for the residual unit from the current unit is possible.

EFFECT: efficient encoding of converted residual moving image signals.

6 cl, 7 dwg

The technical field to which the invention relates.

The present invention relates to a device and to a method of image processing, and, in particular, to a device for reducing the amount of residual image signals and for entropy encoding the residual signal.

The level of technology

For efficient signal transmission of moving images at low data transmission speed with high image quality have been proposed various digital compression of moving images. These technologies compress moving images include standard H.261, MPEG-2/H.262, developed by the Expert group on the moving images (MPEG)standard H.263, MPEG-4, standard Advanced encoding video (AVC)/H.264, etc. compression Technologies include a diagram of the discrete cosine transform (DCT)scheme, motion compensation (KD), the scheme of quantization, entropy encoding scheme, etc.

For coding of images, each image is divided into many layers, and each layer is divided into many blocks of the encoding of a given size. Since the high definition image quality (HD) or a higher class has a lot of planes, the image compression can be improved by encoding image blocks Kodirov is of larger size, than the macroblock MB.

Therefore, since the size of the coding block increases to increase the compression ratio for the image without increasing the complexity of image compression, we need a new way to predict, and the necessary changes in coding with transform entropy coding intra-prediction and inter-prediction.

Disclosure of inventions

Technical task

The present invention relates to a device for encoding a moving image, and, in particular, to a device for efficient encoding of the transformed residual signals of the moving image having a given or larger size.

Technical solution

According to one of the objects of the present invention, an apparatus for encoding a moving image, comprising: a device for determining the encoding mode, configured to divide the input image into unit coding elements and to determine the mode of the prediction unit of the encoding element and the size of the block coding with prediction of this single element encoding; module conversion/quantization configured for transforming and quantizing the residual block between the block coding with prediction and block the predictions generated by intra-before the punishment or inter-prediction; module inverse quantization/transform configured to perform inverse quantization and inverse transform of the quantized transform block into the frequency domain; deblocare filter configured to apply the method deblokiruyuschee filtering, data about the image, taken from the device decoding inverse quantization/transform; generator block prediction, configured to generate a block coding with prediction; and an entropy encoder configured to encode the quantized transform coefficients from the encoder with a transform/quantization, and when the size of the transform block is equal to or greater than the specified size, the entropy encoder divides the quantized transform coefficients into multiple subsets and performs scanning and entropy encoding is not equal to zero quantized transform coefficients of each subset.

Useful effects of the invention

The present invention provides the ability to increase coding efficiency by reducing the amount of residual signal of the block to be coded. Moreover, due to the effective scanning the quantized transform coefficients, except those that are equal to 0, lie in the entropy encoding the number of bits required for entropy encoding is minimized, so that the coding efficiency can be improved.

Brief description of drawings

Figure 1 shows the structure of a partitioning according to a variant implementation of the present invention, which is shown as an example.

Figure 2 illustrates the method of determining the encoding mode according to the present invention.

Figure 3 illustrates a device for encoding a moving image according to the present invention.

In Fig. 4-6 depict the graphs, which illustrate the method of adaptive partitioning according to the present invention.

7 depicts a graph which illustrates the method of encoding the residual signal.

The implementation of the invention

Below is a detailed description of the various embodiments of the present invention with reference to the accompanying drawings. However, the present invention is not limited to the disclosed below its implementation, which are given as examples, but it can be done in different ways. Therefore, there are many other modifications and alterations of the present invention, and it should be understood that within the scope of the disclosed concept, the present invention can be implementing the Vano in practice, otherwise, what, in particular, described above.

For coding of images, each image is divided into many layers, and each layer is divided into a number of individual coding elements of the specified size. Since the high definition image quality (HD) or a higher class has a lot of planes, the compression ratio of the image may be enhanced by the image encoding unit coding elements that are larger than the macroblock MB of size 16×16.

A single element of the encoding according to the present invention may be a block having the size of 32×32, or the block having the size of 64×64, and the macroblocks MB of size 16×16. In addition, a single element of the encoding may be a block having a size of 8×8 or smaller. For convenience, the greatest single element coding is called a super macro block (SMB). The size of the SMB can be determined according to the information indicating the size of the smallest unit of an element coding and information about the depth. Information about the depth indicates the difference between the size of the SMB and the size of the smallest unit of an element encoding.

Thus, a single element of the encoding that will be used to encode all images of a sequence of images may be SMB or sub-block SMB. Valid once the minimum level of individual coding elements can be set by default or in the sequence header. When the allowable size of a single coding elements specified in the sequence header, the allowable size of a single coding elements is set in accordance with the size of the smallest unit of an element coding and information about the depth.

Each image or layer is divided into a number of individual elements SMB. Each SMB or subsection SMB can be encoded by inter-coding or intra-coding and decoded.

To enable accurate decoding unit encoding element (i.e., SMB or sub-blocks SMB) encoder must add information about the size of a coding block with the prediction of a single element coding and information about the prediction mode indicating that a single encoding element was coded bit stream mode intra-prediction or inter-prediction. For this part of the bit stream of single encoding element must be included information on the prediction mode and information indicating the size of the block coding with prediction. The prediction mode is changed in accordance with the layer type.

When the layer type is the type of "intra-" (I), then all the blocks of the coding with prediction in the layer are blocks with intra-prediction, and the prediction block coding with prediction can be determined in the accordance with the block size coding with prediction. However, when the layer type is the type with unidirectional prediction" (P) or bidirectional prediction (B), the prediction block coding with prediction may be determined in accordance with the information on the prediction mode and the block size coding with prediction. Thus, it is preferable to generate the type of a prediction unit of the coding with prediction based on the type layer, information about the prediction mode and information indicating the size of the block coding with prediction, and paste data generated type prediction in the header unit of the encoding element.

When the block coding with prediction coded way of intra-coding, the decoder must provide information about the mode of intra prediction used for intra-prediction, and the type of prediction.

When the block coding with prediction coded way inter-coding, block coding with prediction encoded by any of the following ways: unidirectional prediction and bidirectional prediction. In the case of unidirectional prediction, the bit stream must include information about the reference image and the information about the motion vector used for prediction as well as information about the type of a prediction unit codiovan is predicting for unidirectional prediction. In the case of bidirectional prediction, the header block coding with prediction should include information on two reference images and information about the motion vector used for bidirectional prediction, as well as information about the type of a prediction unit for bi-directional prediction. Information about the motion vector may include information indicating a residual motion vector and the predictor motion vector.

Figure 1 shows the hierarchical structure of separation, illustrating allowable block coding with prediction for SMB, having the size of 64×64, according to a variant implementation of the present invention, which is shown as an example.

When a single element coding using SMB, it is preferable to have four stages of separation into sub-blocks, as shown in figure 1, but the division into blocks is not limited to four stages of division into sub-blocks. If there are four stages of division into sub-blocks, that can be set only 13 types of blocks predictions(64×64, 64×32, 32×64, 32×32, 32×16, 16×32, 16×16, 16×8, 8×16, 8×8, 8×4, 4×8 and 4×4).

There may not be any data to be reported for blocks of inter-coding prediction that is larger than the size of the MB. Thus, it is preferable to re-add mode MB64_SKIP (PROPUSK), when the size of the block coding with prediction equal to 64×64, and the mode MB32_SKIP (PROPUSK), when the size of the block coding with prediction is equal to 32×32. For information transfer mode in the decoder can be used flag MB64_SKIP_flag or flag MB32_SKIP_flag. When the values of these flags is set to 1, then there are no data to be transmitted on the corresponding block coding with prediction.

When multiple consecutive SMB is not coded, then the flag MB64_SKIP_flag can be inserted only in the first SMB, and can be omitted in subsequent SMB. In this case, information about the number of consecutive missed SMB can be added to a layer or to the first SMB. In particular, when multiple consecutive SMB is not coded, then the flag SMB_SKIP_flag first SMB set equal to 1, and it can also be jointly applied to several subsequent SMB. In this case, the layer can be added information corresponding to the number of SMB, which are not consistently coded (for example, SMB_SKIP_number (__SMB)).

When a single encoding element has a size of 32×32, then, in addition to the existing block having a size 16×16, 16×8, 8×16, 8×8, 8×4, 4×8 or 4×4 as block coding with prediction can be used, the block having the size of 32×32, 32×16 or 16×32.

In the mode of inter-prediction type (Mb32_tye) prediction of single encoding element may specify section 32×32, when the type of the prediction is equal to 0, section 32×16, when the type of the prediction is equal to 1, section 16×32, when the type of the prediction is equal to 2, and section 16×16, when the type of the prediction is equal to 3.

When a single element coding is divided into four single subcell encoding these four single subcell encoding encode and transmit in the order corresponding to the raster scan. In this case, the quantization parameter may be transmitted for each unit of the encoding element, and may be transferred only once in the header of a single super-element encoding when the same quantization parameter applied to all single sub coding. However, when the quantization parameter must be changed in a single subcell coding, it can only be transferred to only the differential value with respect to parameter quantization unit of the encoding element of the upper level or the previous single element encoding the same level.

Every single subcell coding can be separated using the method of the Quad-tree, and the sample coded block (cbp) and the residual ratio can also be transferred using the method of the Quad-tree. When using 1-bit cbp, the cbp value of 1 may indicate that a single alimentazione has at least one coefficient that is different from 0, and a cbp value of 0 may indicate that all coefficients are equal to 0.

Figure 2 illustrates the method of determining the encoding mode according to the present invention, when the size of the SMB is 64×64.

As illustrated in figure 2, when SMB is a block of 64×64, the encoder determines whether it is possible or not to skip a block of 64×64, and determines the mode SKIP) as the encoding mode, when you can skip a block of 64×64. The decoder must be transferred to the flag mb64_skip_flag. When the unit is 64×64 data is available, subject to encoding, but it is not divided into blocks of 32×32, then in the SMB header insert information on the size of the encoding SMB, which is one of the following sizes: 64×64, 64×32 and 32×64, and information about whether SMB encoded by way of intra-coding or inter-coding, and generate a data block SMB using the encoded data.

When the unit is 64×64 data is available, subject to encoding, and it is divided into blocks of 32×32, it is similarly determined whether there are in corresponding blocks of 32×32 data to be coded. When corresponding blocks of 32×32 no data to be coded, in the mode for a block of 32×32 define mode SKIP), and the decoder transmit flag mb32_skip_flag.

However, when the block is 32×32 data is available, subject to encoding, but it is not divided into blocks of 16×16, the header blocks of 32×32 inserts the size of the coding unit 32×32, which is one of the following sizes: 32×32, 32×16, 16×32, and information on whether the blocks of 32×32 is encoded by way of intra-coding or inter-coding, and generate the data block using the encoded data.

When a block of 32×32 data is available, subject to encoding, and it is divided into blocks of 16×16, it is determined whether there is data to be coded in the corresponding blocks of 16×16. When corresponding blocks of 16×16, there is no data to be coded, in the mode for a block of 16×16 define the SKIP mode, and the decoder transmit flag mbl6_skip_flag. On the other hand, when the corresponding blocks of 16×16 data is available, subject to encoding in the header block of 16×16 insert size of the coding block of 16×16, which is one of the following sizes: 16×16, 16×8 and 8×16, and the information indicating whether the block of 16×16 is encoded by way of intra-coding or inter-coding, and generate the data block using the encoded data.

When blocks are used with such a variety of sizes, the block size may vary in accordance with each sequence level, picture level, a level layer, SMB, Il is a single subcell encoding SMB.

Figure 3 illustrates a device for encoding a moving image according to the present invention.

With reference to Figure 3, the device for encoding a moving image according to the present invention includes a device 110 determining the encoding mode, the module 120 transformation/quantization, entropy encoder 130, intra-predictor 140, the compensator 150 motion analyzer 155 motion parameters, the module 160 inverse quantization/transform deblocare filter 170, a storage device 180 for images, the adder 185 and subtractive device 190.

The device 110 determining the encoding mode analyzes the input video signal for dividing the image into a single coding elements and determines a size of the block coding with prediction and the prediction mode for each unit of the encoding element. The device 110 determining the encoding mode also passes the block coding with prediction corresponding to a particular size, in a subtractive device 190.

The module 120 transformation/quantization determines the size conversion unit for converting the residual signal received from a subtractive device 190. The size of the transform block may be equal to or smaller than the block size coding with prediction, but can be specified in any other way the m in the inter-mode. In other words, the size of the transform block may be larger than the size of the block coding with prediction in inter-mode. In this case, the module 120 transformation/quantization takes a lot of residual blocks of subtractive device 190 and generates one conversion unit, consisting of a number of residual blocks. The size of the transform block is equal to or less than the size of a single element encoding. The module 120 transformation/quantization performs two-dimensional (2D) discrete cosine transform (DCT) conversion unit to generate transform coefficients. DCT can be integer DCT.

The module 120 transformation/quantization also determines the value of the quantization step used for quantization of the transform coefficients, and performs the quantization of transform coefficients using a quantization matrix defined in accordance with a certain step size of quantization and encoding mode.

The module 160 inverse quantization/transform performs inverse quantization and inverse transform coefficients quantization, quantized module 120 encoding the transform/quantization, whereby to restore the residual block in the spatial domain of the residual block is transformed into h is now the region.

Deblocare filter 170 receives image data subjected to inverse quantization and inverse transformation, from the device 160 inverse quantization/transform, and performs a filtering procedure to remove the effects of blocking. Procedure deblokiruyuschee filtering can be applied to the boundary between blocks coding with prediction and between blocks conversion. This boundary is the edge of the grid having a given size or greater, and a specific size may be 8×8. Procedure deblokiruyuschee filtering includes the operation of determining the boundaries of the subject filtering operation determine the strength of the filter border applied to the border, the operation of determining whether to apply deblocare filter 170, or not, and the selection operation of the filter applied to the border, when it is determined that it is necessary to apply deblocare filter 170.

Do you need to use deblocare filter 170, determined in accordance with the following conditions: (i) whether the strength of the filter border greater than 0 or not, and (ii) whether the value indicating the difference between the pixels on the boundary of block P and block Q, smaller than the first reference value determined according to the quantization parameter or not.

You might have two or more filters. When the absolute value of diff is STI between the two pixels, adjacent to the boundary of the block is equal to or greater than the second reference value, then opt for a weak filter. The second reference value is determined by the quantization parameter and the strength of filtering of the border.

The storage device 180 for images takes the filtered image from deblokiruyuschee filter 170, and stores the image in the form of single images. The image may be an image frame or image field. The storage device 180 for images has a buffer (not shown)capable of storing multiple images.

The analyzer 155 motion parameters analyzes the motion parameters using at least one reference image stored in the storage device 180 for images, and outputs the index of the reference image representing a reference image and a motion vector.

The compensator 150 motion extracts the block prediction, corresponding to the block to be coded, from the reference image used for motion estimation from the set of reference images stored in the storage device 180 for images, according to the index of the reference image and the motion vector, which is introduced from the analyzer 155 motion parameters, and outputs the extracted block prediction.

Intra-predictor 140 performs intra-before the blocks using the restored values of the pixels in the same image. Intra-predictor 140 receives the current block to be coding with prediction, selects one of a specified number of modes of intra-prediction according to the size of the current block, and performs intra-prediction.

Entropy encoder 130 performs entropy coding of the quantization coefficients, quantization are performed by the module 120 transformation/quantization, information about the movement generated by the analyzer 155 motion parameters, etc. of the Quantized transform coefficients is converted into information about the one-dimensional (1D) transform with quantization of the specified scanning method and subjected to entropy encoding.

Below is a description of how intra-prediction with reference to Figure 3.

First, the intra-predictor 140 receives information about the position and the size of the coding block to be coded, from the device 110 to determine the encoding mode.

Then intra-predictor 140 receives the correct reference pixels to determine the mode of intra prediction of the current block coding with prediction of the mass storage device 180 for images. The reference pixels have already been encoded and reconstructed and are adjacent to the current block coding with prediction (below referred to as the current block). When the current block is at the top of the boundary of the current image, the pixels adjacent to the upper side of the current block are undefined. When the current block is located in the left margin of the current image, the pixels adjacent to the left side of the current block are undefined. Moreover, when the current block is located on the top or on the left edge of the layer, the pixels adjacent to the upper or left side, are undefined.

When there are no pixels adjacent to the left or upper side of the current block, or missing any pixels that have been previously encoded and restored, as mentioned above, the mode of intra prediction of the current block may be determined using only the correct pixels.

However, there may be generated an incorrect reference pixel using pixels adjacent to the current block or with the existing pixel. For example, when the pixels of the block are incorrect, the pixels on the upper side can be generated using one or more existing reference pixels on the left side.

Even in the case when there are pixels on the top or on the left side of the current block, the pixels can be defined as incorrect pixels in accordance with the mode of the encoding block to which the pixels belong. For example, when a block, that is, belong to the pixels adjacent to the upper side of the current block has been encoded by the way inter-coding and restored, these pixels can be defined as incorrect pixels. In this case, the reference pixels may be generated using the reference pixel block with intra-mode.

Then intra-predictor 140 determines the mode of intra prediction of the current block using the reference pixels. The number of modes of intra-prediction depends on the block size.

In accordance with the block size allowed 33, 16 or 2 directional mode and at least one non-directional mode. Non-directional mode may be a mode with a DC component (DC) or planar mode.

Blocks having the same size, can be assigned a different number of modes of intra-prediction. To indicate that allowed different number of modes of intra-prediction, at least one of the following headings: in the sequence header, the header image in the header layer and the header of the unit of the encoding element can be pasted information indicating the number of modes of intra-prediction. Preferably this information is inserted in the sequence header or in the header image.

Then, when the determined mode of intra-forecast is of the current block, generate the block prediction for the current block. Block generate predictions using reference pixels that includes the generated pixel, or using a linear combination of the reference pixels on the basis of the mode of intra prediction of the current block. For example, in a directional mode in a particular direction mode predictions can be generated using reference pixels on the top side of the current block and the reference pixels on the left side of the current block.

The correct reference pixels used for generating block prediction can be filtered reference pixels. The number of filters used to correct the reference pixels may be plural. Besides, can adaptively apply multiple filters in accordance with the size of the current block and the mode of intra prediction.

Then the residual block obtained using the current block and the block of the prediction generated by the intra-predictor 140, encode module 120 transform/quantization and entropy encoder 130.

Meanwhile, carry out some coding mode of intra prediction of the current block. The coding mode of intra prediction can be performed intra-predictor 140, a separate device coding mode of intra-prediction (Napoletano) or entropy encoder 130.

The mode of intra prediction of the current block to encode using modes of intra prediction of the upper block and the left block of the current block.

First get profiles of intra-prediction of the left and upper blocks of the current block. When the number of the upper block is plural, as the upper block of the current block is set at the upper left block having the minimum number of the mode. When the number of left blocks is also plural, as the left block of the current block specify the upper-left block having the minimum number of the mode. When the upper block or a left block are not encoded in the intra mode prediction, the mode of intra prediction of the upper block or a left block can be set DC) (mode number 2).

Then, when the mode of intra prediction of the top or left of the block is equal to or greater than the number of allowed modes of intra prediction of the current block, the mode of intra prediction of the top or left of the unit is converted into one of the modes of intra-prediction, allowed for the current block.

Then, when the mode of intra prediction of the current block is identical to one of the following modes: mode of intra prediction of the left block and mode II andstc the prediction of the upper block, pass a flag indicating that the mode of intra prediction of the current block is identical to one of the following modes: mode of intra prediction of the left block and the mode of intra prediction of the upper block, and a flag indicating one of the modes of intra prediction of the upper and left blocks. In this case, when the left and upper blocks of the current block have the same mode of intra prediction may be transferred only a flag indicating that the mode of intra prediction of the current block is identical to one of the following modes: mode of intra prediction of the left block and the mode of intra prediction of the upper block. Similarly, when only one of the modes of intra prediction of the upper and left blocks is correct and identical to the mode of intra prediction of the current block may be transferred only a flag indicating that the mode of intra prediction of the current block is identical to one of the following modes: mode of intra prediction of the left block and the mode of intra prediction of the upper block.

However, when the mode of intra prediction of the current block is different than the mode of intra prediction of the left and upper blocks, determine whether the mode of intra prediction of the current block is smaller than the number of intra mode prediction of the left and upper blocks.

When the determination is prohibited, both rooms of the mode of intra prediction of the left and upper blocks of the current block is greater than the mode of intra prediction of the current block, as the ultimate mode of intra-prediction determine the mode of intra prediction of the current block. However, when only one of the rooms modes of intra prediction of the left and upper blocks of the current block does not exceed the number of the mode of intra prediction of the current block, as the ultimate mode of intra prediction of the current block are intra-prediction, with the mode number obtained by subtracting 1 from the number mode of intra prediction of the current block. In addition, when none of the rooms modes of intra prediction of the left and upper blocks of the current block does not exceed the number of the mode of intra prediction of the current block, as the ultimate mode of intra prediction of the current block are intra-prediction, with the mode number is obtained by subtracting 2 from the number mode of intra prediction of the current block.

Then coding the final mode of intra prediction of the current block. The final mode of intra prediction of the current block code using various code tables in accordance with the fact whether the upper block of the current block is the same mode of intra prediction of that and left the POC of the current block. Mode of intra prediction of the upper block or a left block of the current block may be modified mode of intra prediction. In other words, the mode of intra prediction of the upper block or a left block of the current block may be a mode of intra-prediction, modified in accordance with the table of correspondence of modes of intra-prediction, valid for the upper block and the left block, a given number of modes of intra-prediction. The specified number may be 9 or 3.

The following is a description of how to encode the quantized transform coefficients performed in the entropy encoder 130 of Fig 3.

Entropy encoding the quantized transform coefficients is performed by using context-adaptive coding variable length codes (CAVLC) or context-adaptive binary arithmetic coding (CABAC). When the block size conversion becomes large, there is a high probability that you will need a large number of bits for scanning and entropy encoding of the coefficients other than 0. Thus, it is preferable to introduce a new scanning method of reducing the number of bits for a conversion unit having a size equal to or greater than the specified size.

First, determine whether to split the kV is notowanie transform coefficients into multiple subsets. It depends on the size of the transform block. In other words, when the size of the transform block is equal to or greater than the specified size, the quantized transform coefficients are divided into many subsets. The specified size can be the size of 8×8 or 16×16. Many subsets consists of one main subset and one or more residual subsets. The main subset is located on the upper left side, including the coefficient DC), and one or more residual subsets cover the area except for the area that covers the main subset.

Then, when the size of the transform block is equal to or greater than the specified size, the quantized transform coefficients are divided into the main subset and one or more residual subsets, and quantized transform coefficients contained in each subset, scan and encode. The quantized transform coefficients in the subset can be scanned using one of the many templates scan. Pattern scan, in which the number of bits to be encoded, becomes minimum can be selected in accordance with the distribution is not equal to zero the peak of the oil quantized transform coefficients in the subset. Many templates scan may include zigzag scanning, scanning and vertical scanning horizontally. In addition, it may include scanning or vertical scanning horizontally according to the standard MPEG-4 Expert group on moving images. When the template scan pass for each subset, then this requires a large number of bits. Therefore, for many subsets can be applied one template scan.

The template scan may be selected adaptively in accordance with the prediction mode and direction intra-prediction. For example, in inter-prediction can be applied only one specific pattern scanning (for example, a zigzag scan) or one of the many templates scan. In the first case there is no need to transmit information about the template scan decoder, but in the latter case, information about the template scan need to be passed to the decoder. In the mode of intra prediction of the template scan may be selected in accordance with the direction intra-prediction. For example, can be used to scan horizontally in the mode of intra prediction of the vertical scanning by the vertical mode of intra prediction of the horizontal zigzagoon scanning mode with a DC component (DC).

Pattern scan, used for the main subset and one or more residual subsets may be specified template. The specified template can be a zig-zag scan. In addition to the zigzag scanning, for subsets can be used to scan horizontally or scanning vertically. Pattern scanning applied to subsets can also be adaptively determined in accordance with the prediction mode and direction intra-prediction. In other words, in the mode of inter-prediction for subsets can apply the template scan. When intra-prediction for subsets can apply the template scan is selected in accordance with the direction intra-prediction.

An effective option would be to scan the quantized transform coefficients in the subset in the opposite direction. In other words, the quantized transform coefficients may be scanned with the last nonzero coefficient of the subset in the opposite direction in accordance with the template scan. Similarly, scan many of the subsets is performed in the opposite direction from the subset that includes the last nonzero coefficient of the block of transformed who I am, the main subset.

For the correct entropy decoding in the decoder, the encoder encodes information that can specify the position of the last nonzero coefficient in the transform block, and information capable of indicating the position of the last nonzero coefficient in each subset, and transmits the coded information to the decoder. This information may specify the position of the last nonzero coefficient in each subset. In addition, this information may be information about the map, consisting of flags indicating whether each of the transform coefficients is equal to 0 or not, and flags indicating whether the nonzero coefficient of the last nonzero coefficient in the block conversion or not. The template scan to generate the map information may be the same as the template scan subsets.

In another example, scanning the quantized transform coefficients of a transform block of quantized transform coefficients of the transform block can be reassembled using way interleave and converted into many sub-blocks, and each of the sub-blocks may be scanned and encoded.

When the image has a boundary in a particular direction, and ispolzuetsyaprintsip mode prediction, for similar data on border movement uses different sub-blocks, and can be overhead to the transmission of service information. In this case, it is more effective option would be to split a single item coding in a specific direction in accordance with the shape of the border of the image and the estimation of motion parameters in each divided region.

With reference to Fig. 4-6 will be described an adaptive way in the direction of separation of the block based on the image characteristics. In Fig. 4, 5 and 6, an example will be described in a single encoding element size 32×32. However, the size of a single element coding is not limited to the size of 32×32, and the method can also be applied to a block of 64×64 or for a block of 16×16.

In one of the examples of the simplest adaptive mode single element encoding divide a straight line into two blocks to extract the statistical dependence of field predictions from the local topography. In other words, the border of the image set in compliance with straight lines and share.

As shown in the drawings, when there is a line which passes through the image block is 32×32, then the border, passing through the image should be divided into small blocks for efficient image compression is according to the usual method of dividing the image.

Thus, as shown in figure 4, one block of 32×32 should be separated at least 10 blocks and encode. Therefore, the decoder must be submitted 10 the motion vectors together with information showing the separation of the image, and therefore, required a lot of additional information in addition to information about the image.

As shown in Fig. 5 or 6, when there is a line which passes through the image block is 32×32, the number of servings for more information, transferable in the decoder, can be significantly reduced by matching the boundaries of the image, at least one straight line indicating the border of the image.

For example, when the border, which passes through a block of 32×32, corresponds to the two straight lines, as shown in Figure 5, the block of 32×32 is divided into four blocks of 16×16, and can be obtained straight lines that map to the corresponding first and fourth blocks of 16×16 block borders. In this case, the required six areas of separation, and the number of motion vectors to be transmitted to the decoder may be reduced to six.

Similarly, when the border, which passes through the block that corresponds to a straight line, as shown in Fig.6, the unit of 32×32 is divided into two blocks, and the decoder is necessary is IMO to send only one piece of information about the mode of the block and two motion vectors.

When the unit is divided using a straight line, a decoder is necessary to send information about used straight line. Description information about a straight line, transferable, below.

Information about a straight line, transferable, can be transmitted using various methods.

The first example is the representation of information about a straight line using distance and angle in relation to a specified position. In this case, the predetermined position may be a pixel in the upper left corner or the center of the block. Distance can be an integer or a quantized value. The angle may be an angle between 0 and 180 degrees, and can also be represented by a quantized value.

A second example is the transfer of decoder values of the positions of both ends of a straight line passing through the block. These values of the positions can be expressed as values indicating how far both of these end distant from the pixel in the upper left corner of the block when crawling the boundaries of the block in a clockwise direction, starting from the upper left corner. In this case, either end can be represented as a position relative to the pixel while traversing the boundary in the clockwise direction, and the other can be represented as a position relative to the pixel of the ri crawling boundary in the counterclockwise direction, thus, can be expressed information about a straight line. In this case, information about straight lines that are close to more diverse forms, can be expressed using a small number of bits.

When there are neighboring blocks that have already been adaptive separated by direction and encoded effective option is to transfer the residual information between the information about a straight line in the current block and information on a straight line in one block of the neighboring blocks. The preferred option is to encode the residual information using information about the direction of the blocks divided in the same direction as the current block. Information about the straight line or the residual information of the current block can be expressed in the form of one of the corresponding index, which can be encoded variable length codes and transmitted.

Determining whether to apply adaptive in the direction of separation or not, can be performed in accordance with the size of the block coding with prediction. Because adaptive in the direction of separation, applied to an excessively small block coding can increase the amount of information transferable, and complexity, it is preferable not to use this method.

As shown the and 7, the method of separation of block coding with prediction in accordance with the shape of the border of the image may be applied only to specific areas, which are limited to a specified number of them. For example, the separation unit may be limited to four areas, which are: direction horizontal direction vertical direction diagonally upward and diagonally downward, or two directions, which are: direction horizontally and vertically. A number of variants of the separation unit in a specific direction may be changed in accordance with the size of the block coding with prediction. For example, block coding with prediction, have a size of 32×32, can be divided in a particular direction (e.g., in the direction of the horizontally) using seven methods, and the block coding with prediction, have a size of 16×16, can be divided in three ways. Moreover, regardless of the size of the coding block to the prediction block coding with prediction can be divided using the same number of ways.

In accordance with the chroma block coding with prediction can also be divided into the same sections, and for the separation can also be applied all description is by following methods. Block coding with prediction, with appropriate separation projects should include, in its header, a flag indicating the presence of a separation unit, information indicating how the separation was performed, and the encoded indices of the reference images that are referenced in accordance with the relevant areas of separation.

When one block is predicted by the prediction, adaptive direction, it is necessary to perform the estimation of motion parameters and motion compensation for each of the two divided regions. Thus, from each of the divided areas must be obtained motion vector, and must be received and encoded residual signal between each of the divided areas and the reference area based on the motion vector.

The residual signal may be encoded using any of the following methods.

First, from each of the two separated regions of a single block coding with prediction can be obtained residual signal, and then these two residual signals can be summed to form a single residual signal for the block and encoded with the conversion. In this case, it is very likely the presence of a difference between the General distribution of residual signals in the respective areas divided along the gr is nice, and, therefore, it is preferable to apply a filter to the border.

In another method, the encoding can be performed by applying adaptive to the shape of the transformation to each of the divided areas. As shown in Fig.7, when one block is divided into two areas, the upper left block is subjected to one-dimensional (1D) transform horizontally "as is", and then one-dimensional (1D) transform vertically, and the bottom right block reflow or turn 180 degrees, as shown in the drawing, and subjected to one-dimensional (1D) transform, and then a one-dimensional (1D) transform in the vertical direction. In this case, the residual coefficients, separately encoded according to the corresponding areas of separation, can be transmitted to the decoder, or may be combined and transmitted.

In yet another method for generating and coding block can be done filling insignificant information according to the respective divided regions in Other words, when encoding the current scope split another area of separation, forming a block, fill with values from the current scope of the division for the education of one block, and then subjected to two-dimensional (2D) encoding conversion. Filling may be filling horizontally (up undefined region of the specific region is STI horizontally), and fill vertically (up the undefined region of a certain area vertically). In this case, it is preferable to perform the filling horizontally, and then fill vertically. Moreover, the uncertain pixel is adjacent to one or more specific pixels can be filled with a linear combination of certain pixels. In addition, can be specified direction according to the direction of separation, so that any of the fillings: fill horizontally and fill vertically, may be performed first.

Below is a description of the estimation of the motion vector.

When one block is divided into two regions using a straight line, the motion vector of each of the divided regions encode differently, using already encoded motion vector.

In the first method as the predictor motion vector for the first area of the divided areas can be selected one of the motion vectors of the adjacent blocks, and as a predictor motion vector for the second region can be selected one of the motion vectors of the neighboring blocks, the first region.

In the second method as the predictor motion vector for the first area of the divided areas can be selected one of the motion vectors of the adjacent blocks, and as a predictor motion vector for the second region may be selected motion vector is C the first region.

In the third method, the presence among blocks adjacent to the current block, block, adaptive split in direction, as a reference motion vector using the motion vector of the block adaptive split in direction, taking into account the orientation of the divided blocks. Thus, when there are many blocks, adaptive split in the direction, as a reference motion vectors can be used for motion vectors in the specified sequence or the motion vectors of the blocks having the same direction of division.

In the fourth way to the first area of the divided areas as a predictor motion vector of the first region may be set to the motion vector of a block adjacent to the first area and the second area as a predictor motion vector can be selected any one of the motion vectors of blocks adjacent to the second region, and a motion vector of a block or area of separation in the same location in the previous image, and can be encoded vector residual motion.

Here for partitioning used one straight line. However, one unit can also be divided at least into two regions split using information comprising at least two straight lines, and encoding the divided region which can be performed as described above.

Although the invention has been shown and described with reference to some variations in its implementation, which are given as examples for professionals in the art will understand that can be made various changes in form and detail, without going beyond the nature and scope of the invention defined by the attached claims.

1. Device for encoding a moving image, comprising:
the device determining the encoding mode, configured to divide the input image into unit coding elements and to determine the mode of the prediction unit of the encoding element and the size of the block coding with prediction of this single element encoding;
intra-predictor configured to generate a block prediction using intra-prediction;
module conversion/quantization configured to transform and quantization block conversion of the residual signals obtained using block coding with prediction and block prediction; and
entropy encoder configured to entropy encode the quantized transform block,
moreover, if the size of the quantized transform block is equal to or greater than the specified size of this quantized transform block is divided into multiple subsets and scan the quantized transform coefficients of each subset using a pattern scanning and entropy the encoding.

2. The device under item 1, and the quantized transform coefficients of each subset is scanned in the opposite direction.

3. The device under item 1, with a template to scan multiple subsets is the same as the template to scan the quantized transform coefficients of each subset.

4. The device under item 1, with a template to scan the quantized transform coefficients selected in accordance with the mode of intra prediction of the block coding with prediction.

5. The device under item 1, and the entropy encoder performs the encoding of the position of the last nonzero coefficient of the transform block.

6. The device under item 1, and the set size is the size of 8×8.