KR20070011050A - Method for modeling coding information of a video signal to compress/decompress the information - Google Patents

Abstract

A modeling method for compressing/decompressing coding information of video signals is provided to adaptively perform modeling for binarization using coding information correlated to the coding information of the video signal to improve a data compression rate. A modeling method includes a step of selecting one of a plurality of models to be applied to binarization or inverse-binarization of coding information about an image block on the basis of coding information of a co-located block, which has correlation with the coding information about the image block. The coding information of the co-located block having correlation with the coding information about the image block is identical to the coding information of the image block. The co-located block is adjacent to the left and top of the image block on an image frame decoded prior to the current block.

Description

Method for modeling coding information of a video signal to compress / decompress the information}

1 illustrates a configuration block of a CABAC execution unit in an encoder for performing context modeling according to the present invention.

2 illustrates an example of modeling for adaptively binarizing an input coding information according to the present invention.

3 schematically illustrates a context modeling method for probabilistic (arithmetic) coding of arbitrary coding information.

4 is a block diagram of a CABAC execution unit in a decoder for performing context modeling according to the present invention.

<Description of Symbols for Main Parts of Drawings>

100, 102: context modeler 101: binarization unit

103: general coding engine 104: bypass coding engine

110: Arithmetic coder 201: Inverse binarization unit

202: Context Modeler 203: Generic Decoding Engine

204: bypass decoding engine 210: arithmetic decoder

The present invention relates to a method of modeling information for compressing coding information of a video signal or for decompressing compressed information.

The scalable video codec (SVC) method encodes a video signal at the highest quality and decodes a partial sequence of the resulting picture sequence (a sequence of intermittently selected frames throughout the sequence). Even if it is used, it is a way to enable a low-quality video representation.

SVC is an extended codec from AVC (Advanced Video Codec: 'H.264'), and is one of the entropy coding schemes proposed for use in compressing video signals coded in AVC. Adaptive Binary Arithmatic Coding) can be used for data compression.

CABAC binarizes coding information that is a specific syntax element according to a specified scheme, and then uses the same coding information of a neighboring macroblock to perform modeling for probability coding on the binarized bits. do. When proper modeling is performed according to the probability of each bit value, the compression rate is increased by subsequent arithmetic coding.

However, in addition to modeling for probability coding, the binarization method also affects the compression rate. For example, if a particular syntax element has values in six cases and assigns a bit string such as 0, 100, 101, 110000, 110001, 110010 for those six cases, For example, a small bit number of bits, i.e., zero, is allocated to increase the compression rate.

However, in the coding of an arbitrary macroblock, the case where a specific syntax element has a high probability of occurrence may change. Nevertheless, given a syntax element, the method of binarization is uniformly specified for the syntax element, so that the best compression ratio cannot be obtained in the arithmetic coding at the later stage.

Accordingly, an object of the present invention is to provide a method for adaptively performing a model for binarization by using coding information correlated with the coding information in binarizing coding information to improve data compression ratio.

Another object of the present invention is to provide a method of using modeling information applied to binarization of coding information as a basis for modeling for probability coding on a bit string of the coding information.

Another object of the present invention is to provide an adaptive modeling method applied to release compressed coding information in response to adaptive modeling during encoding.

The modeling method according to the present invention for achieving the above object, based on the coding information of the adjacent block correlated with the coding information for the coding information for the image block, the binarization or inverse of the coding information And selecting one model from a plurality of models to be applied to inverse-binarization, wherein the plurality of models are distinguished from each other according to coding information of the adjacent block.

In one embodiment according to the present invention, the coding information of an adjacent block correlated with the coding information is the same coding information as that of the video block.

In one embodiment according to the invention, the neighboring block is a block adjacent to the left and top of the video frame, which is decoded before the current block.

In one embodiment according to the present invention, each model in the plurality of models is configured to allocate a small number of bits for a case where the probability that the coding information of the image block has high according to the coding information of the neighboring block. It has a bit conversion structure.

In one embodiment according to the present invention, the model information selected for the binarization is used to determine a model for probability coding the coding information for the binarized image block.

Hereinafter, with reference to the accompanying drawings, a preferred embodiment of the present invention will be described in detail.

1 is a block diagram of a CABAC execution unit according to an embodiment of the present invention. The configuration of FIG. 1 includes a first context modeler 100 that performs modeling for binarization based on coding information 10 of a neighboring macroblock with respect to inputted corresponding coding information. The binarization unit 101 for binarizing the input coding information according to a method specified by the model determined by the first context modeler 100, and coding of an adjacent block for each bit of the binarized coding information. Arithmetic coding of input bits based on information and / or a second context modeler 102 modeling based on information 11 modeled by the first context modeler 100 It includes an arithmetic coder 110.

The arithmetic coder 110 includes a regular coding engine 103 for arithmetically coding each bit of coding information based on variables (probability function and initial value) modeled by the second context modeler 102. It includes internally the bypass coding engine 104 which performs arithmetic coding on the coding information itself, which has no advantage of modeling since the probability of occurrence of bits 0 and 1 is almost equal.

The present invention relates to modeling of input coding information, and the arithmetic coding process according to the modeled variable is not directly related to the present invention. Therefore, in the following description, a description of the bit compression operation (entropy coding) related to the arithmetic coder 110, which is not necessary to aid the understanding of the present invention, will be omitted.

When the input coding information is not binarized, the CABAC execution unit of FIG. 1 sets a model for binarization to be performed on the coding information in the first context modeler 100. One example of modeling is shown in FIG. 2.

The coding information used in the example of FIG. 2, that is, the syntax element is for a macro block type (mb_type), and a model to allocate a bit string for each type (Direct, Intra, P_16x16, P_16x8, P_8x16, and P_8x8) is 5 Each model is distinguished from each other according to the macroblock types of the adjacent blocks A and B of the block X of the macroblock type to be currently modeled. Each model presented separately in the example of FIG. 2 is merely to show that the model to be allocated to the macroblock type of the current block (X) varies according to the macroblock type of the neighboring block. The number of models can be reduced by further subdividing the case according to the macro block type of) or by adding more models or by dividing the case more comprehensively.

In the example of FIG. 2, if the adjacent blocks A and B are Intra mode, it can be predicted that the current block X is also likely to be Intra mode. This allocated model 2a is selected, and if the adjacent blocks A and B are in Direct or P_16x16 mode, it is predicted that the probability that the current block X is the intra mode is the lowest, so that the number of bits for the intra mode is the most. It shows that the model (2b or 2c) that has the most and most bit changes is selected.

The scheme allocated for each case of the macroblock type in one model, as shown in FIG. 2, has a probability that the type of the current macroblock X will have according to the type of the macroblocks A and B of the adjacent block. It is only an example of arrangement in the case of being estimated to be high, and the bit string allocation order may be changed in a different order from the order within each model shown in FIG.

When a model is selected by the first context modeler 100, information about the selected model is provided to the second context modeler 102, and the binarization unit 101 is input according to the selected model. Assign binarization values to macro block types.

In the above description, although the macroblock type is taken as the coding information, the method of adaptively selecting a model for binarization can also be applied to other coding information based on the coding information of another adjacent block. In addition, for any coding information, instead of using only the same coding information of the adjacent block as the basis of the binarization model selection, other coding information correlated with the coding information of the current block may be used as the basis of the binarization model selection.

The binarized bit string as described above is applied to the arithmetic coder 110 at the rear end for bit compression. At this time, a bit having a probability that the value of the bit is evenly distributed without biasing to either one of 0 and 1 A bit of coding information which is directly applied to the bypass arithmetic coder 104 and has a probability of biasing to either one of 0 and 1 is applied to the second context modeler 102 to undergo a modeling process first.

The second context modeler 102 performs modeling on the applied bit string based on the same coding information of the adjacent macro block and the information 11 modeled by the first context modeler 100. . Modeling determines the selection of a probability function and an initial value in the probability function. As illustrated in FIG. 3, an offset value (, k-1, k, k + 1,) is determined according to the type of coding information. And a probability function (, f _k-1 , f _k , f _{k + 1} ,) is first selected, and its offset is determined according to the same coding information of the adjacent block and the information 11 modeled by the first context modeler 100. The value of the index variable (ctxIdxInc) is determined from the value.

When the second context modeler 102 models arbitrary coding information, when the same coding information of an adjacent block is divided into the same number of cases as the first context modeler 100, the coding of the adjacent block is directly performed. The value of the index variable may be determined by referring only to the modeling information 11 provided from the first context modeler 100 without referring to the information. However, when the first context modeler 100 and the second context modeler 102 classify the coding information of adjacent blocks into different cases in modeling the same coding information, for example, the first context modeler 100 and the second context modeler 102. The context modeler 100 reduces the number of models by roughly classifying the number of cases according to coding information of an adjacent block, and the second context modeler 102 models the number of cases according to coding information of an adjacent block in more detail. You can have more. In this case, the second context modeler 102 determines the index variable ctxIdxInc by referring to both the information 11 modeled by the first context modeler 100 and the coding information of the adjacent block.

Alternatively, the offset value is determined according to the coding information and the value of the information 11 modeled by the first context modeler 100 to select a probability function, and according to the coding information of the adjacent block having correlation with the coding information. The value of the index variable (ctxIdxInc) may be determined from the determined offset value.

When the index variable is determined, an initial value (valMPS, pStateIdx) to be applied to the probability function is determined (as shown in FIG. 3, when an initial value of pStateIdx is determined, an initial probability value of LPS (or MPS) is determined accordingly). In this case, the general coding engine 103 of the rear stage encodes (compresses) the input bits using the determined probability function and both initial values (valMPS, pStateIdx) as starting points.

The above description is based on the compression process of coding information in the encoder, but the modeling method described above may be equally applied to the CABAC decoder of FIG. 4 for releasing the compressed data.

The context modeler 202 of the CABAC decoder of FIG. 4 also uses the neighboring block of the adjacent block in the same manner as the second context modeler 102 in the CABAC coder of FIG. 1 employs in modeling for probability coding. The initial value is passed to the regular decoding engine 203 at a later stage by modeling based on the correlated coding information 20. When modeling, the binarization model of the current block according to the coding information of the adjacent block is also grasped and the information 21 of the identified model is transmitted to the inverse binarization unit 201 at the rear end. Of course, a separate context modeler for identifying the binarization model required for inverse binarization may be disposed at the front end of the inverse binarization unit 201. In addition, the coding information of the adjacent block is inversely binarized by the inverse binarization unit 201 prior to the current block, so that the context modeler 202 may refer to the determination of the binarization model of the current block. Delivered (22).

The general decoding engine 203 converts the compressed bits inputted from the same value as the initial value starting from the general coding engine 103 of the encoder for bits of arbitrary coding information into an uncompressed bit string. The inverse binarization unit 201 checks the value of the corresponding coding information with respect to the transformed bit string based on the received information about the binarization model 21. That is, in the modeling example of the macroblock type of FIG. 2, when the decompressed bit string is "100", when the information 21 about the binarization model delivered by the context modeler 202 is a value for model 2a, In P_8x8 mode, the value for model 2b is P_16x16 mode, the value for model 2c is Direct mode, the value for model 2d is P_8x16 mode, and the value for model 2e is P_16x8.

The decoder including a context modeler for modeling coding information according to the above-described method may be mounted in a mobile communication terminal or the like or in an apparatus for reproducing a recording medium.

It is to be understood that the present invention is not limited to the above-described exemplary preferred embodiments, but may be embodied in various ways without departing from the spirit and scope of the present invention. If you grow up, you can easily understand. If the implementation by such improvement, change, replacement or addition falls within the scope of the appended claims, the technical idea should also be regarded as belonging to the present invention.

As described above in a limited embodiment, the present invention adaptively performs a binarization model for the current coding information based on the coding information of the adjacent block, thereby reducing the number of bits of bits in the value of the probabilistic high coding information. Allow the column to be allocated. In addition, this adaptive binarization model makes the probability that the coding information is biased to a specific bit value (0 or 1) is higher than that of the uniform bit allocation method. This further improves the compression ratio by probability coding performed after binarization.

Claims (8)

In the method for modeling to compress the coding information of the video signal, Selecting one model from a plurality of models to be applied to binarization of the coding information, based on the coding information of an adjacent block correlated with the coding information, with respect to the coding information of the image block. It's done, And the plurality of models are distinguished from each other according to the case of coding information of the adjacent block. The method of claim 1, And coding information of an adjacent block correlated with the coding information is the same coding information as that of the video block. The method of claim 1, Determining a model for arithmetic coding the coding information for the binarized image block based on the selected model; And using the model information selected for the binarization as a basis for model determination for the arithmetic coding. The method of claim 1, Wherein each model in the plurality of models is a model that allows a small number of bits to be allocated for a case where the coding information of the image block has a high probability according to the coding information of the neighboring block. A method for releasing coding information of a compressed video signal, For the binarized coding information for an image block, one model is selected from a plurality of models to be applied to inverse-binarization of the coding information based on the coding information of an adjacent block correlated with the coding information. Including, but not limited to, And the plurality of models are distinguished from each other according to the case of coding information of the adjacent block. The method of claim 5, And coding information of an adjacent block correlated with the coding information is the same coding information as that of the video block. The method of claim 5, The adjacent block is a block in which the coding information is inversely binarized prior to the video block. The method of claim 5, Wherein each model in the plurality of models is a model that allows a small number of bits to be allocated for a case where the coding information of the image block has a high probability according to the coding information of the neighboring block.

Images