KR20010069016A - An Itra/Inter Coding Mode Decision Method For Video Coding - Google Patents

Abstract

PURPOSE: A decision method of an intra/inter coding mode in a video encoder is provided to optimize a coding efficiency, and to improve a coding ability and an image quality of the video encoder by deciding the intra/inter coding mode according to the number of bit used in a coding work of a quantization parameter and a motion vector. CONSTITUTION: Firstly, a video signal of block unit is inputted for a block-based animation coding. Next, the variance information(VAR(C_Block)), a residual signal energy(SAD(C_Block)), and motion vector difference information(delta MV) are calculated. At this time, the variance information(VAR(C_Block)), the residual signal energy(SAD(C_Block)), and the motion vector difference information(delta MV) are for considering the number of bit used in a coding work of a quantization parameter(QP) and a motion vector(MV) of a current block. Finally, a coding mode of the current block is selected between an intra mode and an inter mode by comparing the variance information(VAR(C_Block)), the residual signal energy(SAD(C_Block)), and the motion vector difference information(delta MV). At this time, when the variance information(VAR(C_Block)) is bigger than the sum of the residual signal energy(SAD(C_Block)) and the result of multiplication of gamma, the quantization parameter(QP), and the sum of an absolute values of components of the motion vector, the coding mode of the current block is the inter mode.

Description

An Itra / Inter Coding Mode Decision Method For Video Coding}

The present invention relates to an efficient method for determining an intra mode and an inter mode in video encoding. In particular, when block-based video encoding is performed, a displacement amount (deviation) of a current block 'VAR (C_Block) is used. ), The SAD (Sum Absolute Difference) 'SAD (C_Block)' of the current block, and the quantization parameter (QP) and the difference between the motion vector (ΔMV), and constants related to bit allocation and residual signal coding (β, The present invention relates to a method of determining an intra / inter encoding mode of an image encoder by considering a coding mode in consideration of K _T ).

Various techniques and standards for compressing video or still images have been proposed, and the standards for compressing video and still images have the ultimate purpose of compressing and encoding video with more efficient and higher performance.

Currently, MPEG-1, MPEG-2, MPEG-4, and H.261, H.263, H.26L, etc. have been proposed or proposed (adopted) as standards related to video encoding of moving pictures and still images.

Like many international standards, the H.26L standard performs motion estimation and encodes the residual signal resulting from spatio-temporal transformation and zigzag scanning. ) May be performed based on encoding by a quantization process.

In such image encoding, an image is encoded on a block basis. For example, a spatio-temporal transform (for example, a DCT transform, etc.) is performed on an INPUT, quantized and transformed. The compression coefficients are outputted by performing the variable length coding (VLC) by scanning the transform coefficients using a suitable technique such as single scan or double scan.

1 is a diagram illustrating a configuration of a general video encoder.

The input video signal is discrete cosine transformed by the discrete cosine transform unit (DCT) 102 through the encoding mode determination step 101.

The DCT coefficients passed through the DCT are quantized by the quantization unit (Q) 103 and compressed and output together with a motion vector (MV) by the entropy coding unit 104 such as variable length coding (VLC).

In addition, the quantized signal is restored through the inverse quantization unit (Q ⁻¹ ) 105 and the inverse DCT unit (DCT ⁻¹ ) 106.

The reconstructed signal is stored in the image memory 108 together with the signal of the motion estimation and compensation unit 107, and the motion estimation and compensation unit 107 reconstructs the input image signal and the memorized image signal (previous frame / block). Motion estimation and compensation are performed together.

That is, as shown in FIG. 1, the video encoder (encoding method) is a process such as preprocessing of the image, motion estimation and compensation, transform, quantization, variable length coding (VLC), determination of encoding mode, and rate control. In almost all block-based video encodings that have been made or have been standardized, whether to encode in macro mode for each macroblock (Macro Blcok) as shown in FIG. , Or whether to encode in inter mode.

In intra mode, macroblocks are generally encoded through transform, quantization, and variable length encoding, in which case motion estimation and compensation are not needed.

On the other hand, in the inter mode, motion estimation is performed on the current block, and the motion vector is obtained as the displacement of the predicted block with respect to the current block.

After motion compensation from this motion vector (MV), subblocks (i.e., residual blocks) are made by subtracting pixel values of the predicted block from the current block, and then encoding is performed in the same way as intra macroblocks for the subblocks. The motion vector (MV) and texture information of the difference block are transmitted to the decoder.

With this information, the decoder can obtain a reconstructed image of the current block by adding the quantized difference block to the predicted block by the motion vector.

However, in general, since a moving picture with little motion has a large time correlation, the residual signal of the difference block between each block and the predicted block in the current frame is very small.

In this case, the number of bits for encoding the residual signal and the number of bits for encoding the motion vector are smaller than the bits for encoding the current block.

On the other hand, when overlapping or exposed areas are generated and scene transitions are caused by rapid movement or movement of an object, prediction of a macro block generally fails, and in this case, the number of bits used to encode a motion vector and a residual signal is increased. Since the number of bits used to encode a block is greater than that, it is selected in the intra mode, and it is preferable to encode the original image in terms of encoding efficiency and subjective quality.

Currently, the following parameters are used in H.263 to determine the intra / inter coding mode.

(a). MB_mean (macro block average), (b). The difference between the original image and the macroblock mean (Variance A), (c). Difference information SAD (x, y) (SAD: Sum Absolute Difference) between the original image and the encoded previous image.

Here, the macro block average MB_mean is obtained by dividing the total of 16 × 16 pixels (one macro block) of the original image by '256' to obtain an average value of the macro block, and the displacement amount A is obtained from the original image. SAD (x, y) is the SAD value of the motion vectors x and y having the smallest SAD value, and the corresponding block of the original image. The sum of the difference of the corresponding blocks of the previous coded image can be obtained.

In consideration of the above parameters, the intra mode is selected when the displacement A of the original image satisfies the following condition, that is, A <SAD (x, y) -500, otherwise the inter mode is selected.

Are those similar to the MPEG4 using 2NB instead of '500', A <SAD ( x, y) - 2NB (NB if the number of pixels of the macroblock that is, the 16 × 16 macroblock ² × 16 2 = 512) Intra mode is selected when is satisfied, otherwise inter mode is selected.

As described above, in the determination of the intra mode and the inter mode, the encoding mode is determined by simply comparing the energy of the predicted residual signal and the energy of the original image. In this case, the number of bits according to the quantization parameter QP is determined. We do not consider the effect of and the number of bits used to encode the motion vector.

Therefore, it is difficult to expect the optimum coding efficiency because only the comparison of the SAD, which is the energy of the predicted residual signal, with the displacement value of the pixels of the macroblock, does not consider the number of bits used to encode the quantization parameter and the motion vector.

In the block-based video encoding, the present invention determines the optimal intra / inter encoding mode by determining the intra / inter encoding mode in consideration of the number of bits required for encoding a quantization parameter (QP) and a motion vector. The present invention provides a method of determining an intra / inter encoding mode of an image encoder, which has encoding efficiency and can improve encoding performance and image quality of a video encoder.

In particular, the present invention, when performing the block-based video encoding, the displacement (variation) 'VAR (C_Block)' of the current block, the sum Absolute Difference (SAD) 'SAD (C_Block)' and the quantization parameter (QP) and the motion of the current block Image considering the number of bits in encoding the quantization parameter and the motion vector by determining the encoding mode in consideration of the difference (ΔMV) of the vector and the constants (β, K _T ) related to bit allocation and residual signal encoding. A method of determining an intra / inter encoding mode of an encoder is provided.

1 is a block diagram of a video encoder

2 is a diagram for explaining differential coding of a motion vector.

3 is a flowchart of a method for determining an intra / inter encoding mode of the present invention.

A method of determining an intra / inter encoding mode of a video encoder of the present invention; (a). Inputting a video signal for block-based video encoding, (b). Obtaining the displacement amount VAR (C_Block) of the current block, the predicted residual signal energy SAD (C_Blcok), and the displacement (difference) ΔMV of the motion vector, and (c). Determining the intra mode and the inter mode by considering the displacement amount, the energy, and the displacement of the motion vector together with the bit allocation constant β and the residual coding relationship constant K _T for the differential motion vector. A method of determining an intra / inter encoding mode of a video encoder.

In the present invention, when the conditions of VAR (C_Blcok) <SAD (C_Blcok) + γQP | ΔMV ∥ (γ = β / K _T ) is determined as an intra mode, otherwise it is characterized as an inter mode.

A method of determining an intra / inter encoding mode of the video encoder of the present invention as described above will be described in more detail.

In the selection of intra mode and inter mode, the most ideal method is to calculate the number of bits when coding in intra mode when distortion is the same, the number of bits for coding residual signal after motion estimation and compensation, and the motion vector. It calculates the number of bits when encoding and compares them.

In other words, if Bits (VLC (Q (T (C_Block)))) <Bits (VLC (Q (T (MC (C_Block)))) + Bits (MV), it is encoded in intra mode, otherwise it is inter mode. To encode

This means that the number of bits for the result of performing transform (T) and quantization (Q) and variable length coding (VLC) on the current block (C_Blcok), motion compensation (MC) and transform (T) on the current block And if it is smaller than the number of bits for the result of performing the quantization (Q) and the variable length coding (VLC), it means encoding in the intra mode, otherwise encoding in the inter mode.

That is, the number of bits Bits (VLC (Q (T (C_Blcok)))) (that is, Bits (Texture (C_Block))) when encoding in the intra mode is used to encode residual signals after motion estimation and compensation. If the number of bits (Bits (VLC (Q (T (MC (C_Block)))))) (i.e. Bits (Residual (C_Block))) and the number of bits when encoding the motion vector (Bits (MV)) In this case, the encoding mode is selected as the intra mode, and otherwise, the encoding mode is selected as the inter mode.

The transform T may be a DCT, and other types of transforms may be possible.

However, this method will be a technique of actually performing the encoding directly in the intra mode and the inter mode, and then selecting a method that is more efficient (when considering the number of bits). The disadvantage is that it is difficult to clearly define the relationship with numbers mathematically.

Therefore, for simplicity of calculation and processing, Bits (VLC (Q (DCT (C_Block)))) is related to the texture of the current block (C_Blcok), displacement (Variance), quantization parameter (QP), and Bits (VLC ( Q (DCT (MC (C_Block)))) is related to the SAD and quantization parameter QP of the current block C_Blcok.

As described above, when compressing and encoding a moving picture or a still image through block-based motion estimation, each block in the image frame has one motion vector, and there is a degree of correlation between the motion vectors of adjacent blocks.

Therefore, in order to encode a motion vector of each block, a motion vector prediction vector for a block to be currently encoded is obtained using the motion vectors of blocks of adjacent previous frames, and the difference between the motion vector of the adjacent blocks and the motion vector of the adjacent blocks ( Difference Vector encoding is used as a more efficient encoding technique in block-based image encoding.

That is, in MPEG-4 and H.263, as shown in FIG. 2, the middle of three spatially neighboring neighboring motion vectors MV1, MV2, and MV3 previously transmitted in order to predict the current motion vector MVc. Using a value Median, variable length coding is applied to the difference between the actual motion vector and the predicted motion vector.

Therefore, since the number of bits Bits (MV) used to encode the motion vector for determining the encoding mode is differentially encoded with the motion vector of the neighboring block, it is related to the magnitude of ΔMV.

That is, ΔΔMV ∥ = ∥Mcc-Median (MV1, MV2, MV3) ∥ (ΔMV = ΔMVx, ΔMVy)

Here, ΔMVx and ΔMVy consider that the motion vector is a two-dimensional vector consisting of two components, a horizontal direction (x) and a vertical direction (y).

Therefore, ∥ΔMV∥ = ｜ ΔMVx | + | ΔMVy |

Median (MV1, MV2, MV3);

MVx = Median (MV1x, MV2x, MV3x), MVy = Median (MV1y, MV2y, MV3y).

Therefore, the method of determining the intra / inter encoding mode of the video encoder of the present invention;

As described above, this is the most ideal case for intra mode encoding or inter mode encoding determination;

(a). Bits (VLC (Q (T (C_Block))))

(b). Bits (VLC (Q (T (MC (C_Block)))))

(c). Bits (MV)

In order to simplify the calculation and processing,

(a). Bits (VLC (Q (T (C_Block)))) is a function of Bits (Texture (C_Blcok)) → texture displacement amount VAR (C_Blcok) and quantization parameter QP of the current block; K _T / QP (VAR (C_Blcok)) (where K _T is a constant related to bit allocation for differential motion vectors)

(b). Bits (VLC (Q (T (MC (C_Block))))) is a function of Bits (Residual (C_Blcok)) → SAD (SAD (C_Blcok)) and quantization parameter QP of the current block; K _R / QP (SAD (C_Blcok)) (where K _R is a constant related to residual signal coding),

(c). Bits (MV) is a function that considers the magnitude of ΔMV due to differential coding and motion vectors of neighboring blocks; Redefine and approximate as β ΔΔMV as follows.

That is, K _T / QP VAR (C_Blcok) <K _R / QP SAD (C_Blcok) + β ∥ΔMV∥

_{= VAR (C_Block) <K R} / K T · SAD (C_Block) + β / K T · QP · ∥ΔMV∥

In the case of, the signal is encoded in the intra mode. Otherwise, the signal is encoded in the inter mode.

In addition, in consideration of K _T ≒ K _R , K _T / QP ≒ K _R / QP experimentally,

It is possible to redefine the determination condition of the coding mode by VAR (C_Block) &lt; SAD (C_Block) + γ · QP · ∥ΔMV∥ (γ = β / K _T ).

Therefore, the VAR is compared with each other by comparing the sum of the magnitude of the motion vector difference considering the estimated residual signal energy (SAD (C_Block)) and γ with respect to the displacement of the current block (Variance Current Block = VAR (C_Block)) and γ. If (C_Block) is smaller than SAD (C_Block) + γ · QP · ΔMV, the encoding is performed in the intra mode. Otherwise, the encoding is performed in the inter mode.

3 illustrates a method of determining an intra / inter encoding mode of the image encoder of the present invention.

As can be seen from the method of determining the intra / inter mode described above and the number of bits used to encode the vector, the number of bits used to encode the vector is proportional to the quantization parameter QP. As the quantization parameter QP increases, the texture and residual signals are encoded. The ratio of the number of bits of the motion vector to the number of bits to be used becomes large.

However, if VAR (C_Block) &lt; SAD (C_Block) + γ · QP · ∥ΔMV∥ (γ = β / K _T ), the encoding mode determination method of the present invention performs encoding in the intra mode, otherwise the inter mode. Looking at it according to the characteristics of the block can be interpreted as the following four.

First, in the case of overlapping or revealed region generation and scene change caused by rapid movement or object movement, the values of SAD (C_Block) and γ · QP · ∥ΔMV∥ increase, which increases the probability of being selected as an intra block.

Second, when there is small movement, the values of SAD (C_Block) and γ-QP.

Third, if the previous macro block is encoded in the intra mode, the previous motion vector becomes '0' and the value of γQP ∥ΔMV ∥ increases, which increases the probability that the current macro block is encoded in the intra mode. do.

Fourth, since the number of bits used to encode the motion vector is proportional to the quantization parameter QP, as the quantization parameter QP increases, the ratio of the number of motion vector bits to the number of bits used to encode the texture and residual signal increases. As a result, the probability of encoding the intra mode increases.

Therefore, considering these four aspects, the probability of encoding an intra block is determined by the occurrence of regions overlapped or exposed by rapid movement or movement of an object, when a scene change occurs, and when a previous macro block is encoded intra, or quantization. The larger the parameter QP becomes, the greater the probability that the interblock is encoded is well selected when the neighboring blocks are static or have slow motion.

According to the present invention, in block-based video compression encoding, an efficient intra / inter encoding mode can be determined in consideration of the number of bits required for encoding a quantization parameter (QP) and a motion vector (MV). You can expect an improvement in image quality.

In addition, since the present invention can be suitable for encoding of low complexity video, it is advantageous for a video compression encoding and transmission system in real time communication, and is particularly effective for a low rate video transmission technology for a video telephone using a public network (PSTN).

In addition, since the present invention determines a block to be encoded in the intra or inter mode, for an image (images with many exposed or overlapping areas in a scene due to scene change or rapid movement of an object) affected by the mode decision. High efficiency

In addition, in the case of natural video input directly to the camera rather than the MPEG-4 and H.263 test video, the generation of intra macro blocks is more useful in this case.

Claims (6)

In determining the intra encoding mode and the inter encoding mode in block-based video encoding, (a). Inputting a video signal in units of blocks for block-based video encoding, (b). Displacement information (Variance (C_Block)), residual signal energy (Residual SAD), and motion vector difference considering the number of bits for encoding the quantization parameter QP and the motion vector MV with respect to the block C_Block to be currently encoded. Obtaining information [Delta] MV, (c). Comparing the obtained information to determine an intra / inter mode of a block to be currently encoded; A method of determining an intra / inter encoding mode of an image encoder, characterized in that consisting of. The method of claim 1, wherein the displacement information Variance (C_Block) of the current block considering the quantization parameter (QP) is; A method of determining an intra / inter coding mode of an image encoder, characterized in that it is considered as a function of K _T / QP · VAR (C_Block) (K _T is a constant). The method of claim 1, wherein the residual signal energy considering the quantization parameter is Residual SAD; A method of determining the intra / inter coding mode of an image encoder, characterized in that it is considered as a function of K _R / QP SAD (C_Block) (K _R is a constant). The motion vector difference information (ΔMV) according to claim 1, wherein the motion vector difference information (ΔMV), which takes into account the number of bits required to encode the motion vector (MV); A method of determining an intra / inter coding mode of an image encoder, characterized in that it is considered as a function of β ΔΔMV ∥ (β is a constant). The displacement information Variance (C_Block) according to any one of claims 1 to 4, which takes into account the number of bits used to encode the quantization parameter QP and the motion vector MV with respect to the block C_Block to be currently encoded. )) And residual signal energy (Residual SAD) and motion vector difference information (ΔMV) are compared, and if the displacement information is smaller than the residual signal energy and motion vector difference information, encoding is selected as an intra mode. A method of determining an intra / inter encoding mode of an image encoder, characterized in that encoding is selected as a mode. In determining the intra encoding mode and the inter encoding mode in block-based video encoding, (a). Inputting a video signal in units of blocks for block-based video encoding, (b). Displacement information VAR (C_Block) and residual signal energy information SAD (C_Block), and motion, which take into account the number of bits for encoding the quantization parameter QP and the motion vector MV for the block C_Block to be currently encoded. Obtaining the vector difference information ([gamma] QP, [Delta] MV, ([gamma] = [beta] / _KT )), (c) .comparing the information obtained in the above steps, the condition of VAR (C_Blcok) &lt; The method of determining the intra / inter encoding mode of the video encoder, characterized in that the intra mode is determined when it is satisfied, and the inter mode is determined otherwise.