KR100701129B1 - High speed block mode decision method - Google Patents

Abstract

A fast block mode decision method in a video encoder is provided to greatly reduce a calculation quantity by primarily judging whether to perform an inputted macro block as a skip block mode and omitting motion search for an inter mode with respect to the macro block judged as the skip block mode. A video encoder obtains an adaptive threshold value in consideration of motion between neighboring frames(400). The video encoder discriminates whether to perform an inputted 16Î16 macro block as a skip block mode by using the adaptive threshold value(410,420). If it is discriminated that the 16Î16 macro block is performed as the skip block mode, the video encoder obtains coding costs through motion search for the skip block mode(430). If it is discriminated that the 16Î16 macro block is not performed as the skip block mode, the video encoder obtains coding costs for respective inter modes through motion search with respect to the inter modes(440-470). The video encoder obtains coding costs for respective intra modes through motion search with respect to the intra modes(500,510). The video encoder compares the obtained coding costs, and decides a block mode having the minimum coding cost as the last block mode for the motion prediction of the 16Î16 macro block(520).

Description

High speed block mode decision method in image encoder

1 is a block diagram schematically illustrating a block-based image encoder for explaining an operation of compressing an input video signal.

2 is a diagram illustrating various block modes for motion search as in H.264.

3 is a flowchart specifically illustrating a motion search process for determining a conventional block mode.

4 is a flowchart illustrating an embodiment of a fast block mode determination method according to the present invention using the characteristics of the SKIP block mode as described above.

The present invention relates to an image encoder, and more particularly, to a fast block mode determination method for determining a block mode for encoding through motion estimation.

1 is a block diagram schematically illustrating a block-based image encoder for explaining an operation of compressing an input video signal. The image encoder illustrated in FIG. 1 includes a block unitization unit 10, a subtraction unit 20, a block discrete cosine transformer (block DCT, 30), a quantizer 40, a variable length encoder 50, a multiplexer 60, Motion compensator 70, inverse quantization / IDCT 80, motion estimator 110, frame memory 100, and adder 90.

Referring to FIG. 1, the block unitization unit 10 divides every frame of a video input in block units. The subtractor 20 subtracts the frame I (t ') provided by the motion compensator 70 from the current frame I (t) provided by the block unitization unit 10. The block DCT 30 performs discrete cosine transform on the result subtracted by the subtractor 20, and the quantizer 40 quantizes the transformed coefficients in the block DCT 30. The variable length encoder 50 generates a compressed stream by performing variable length coding on the quantized value in the quantizer 40. Meanwhile, the inverse quantizer / IDCT 80 performs inverse quantization and inverse discrete cosine transform on the quantized value in the quantizer 40, and the adder 90 performs a motion compensator 70 inverses the inverse quantization and inverse discrete cosine transform. Is added to the frame provided by the. The frame memory 100 stores the result added by the adder 90 as the previous frame I (t-1). The motion estimator 110 predicts the motion vector MV by comparing pixel values between the previous frame I (t-1) and the current frame I (t) stored in the frame memory 100. The motion compensator 70 generates a new frame I (t ') by applying the motion vector MV provided from the motion predictor 110 to the previous frame I (t-1) provided from the frame memory 100. The multi-plexer 60 multiplexes the image data generated by the variable length encoder 50 or the motion vector MV generated by the motion estimator 110.

Meanwhile, current image compression standards detect a motion vector (MV) by applying a block-based matching algorithm to the motion estimator 110 because of ease of implementation. In this case, data to be transmitted can be reduced by maximizing the correlation between the two frames. H.264, one of the compression standards for video signals, performs motion search on the macro block currently input by the motion estimator in various block modes, and selects the block mode with the highest coding efficiency as the final block mode of the macro block. Therefore, a large gain in coding efficiency is provided. In addition, since a plurality of reference pictures are supported, a block having the greatest overlap with the current block may be selected as a reference picture, thereby obtaining greater coding efficiency than when only the previous frame is used as the reference picture.

2 is a diagram illustrating various block modes for motion search as in H.264. In H.264, among the seven possible block modes such as inter mode (Inter16x16, Inter16x8, Inter8x16, P8x8), intra mode (INT16x16, INT4x4), and SKIP block mode shown in FIG. Determine the block mode of the final macroblock. Here, the coding cost is a rate-distortion optimization function value in consideration of deterioration of image quality and bit amount.

3 is a flowchart specifically illustrating a motion search process for determining a conventional block mode.

Referring to Figures 2 and 3, a conventional block mode determination process will be described. In H.264, when searching for a motion to determine a block mode of a macroblock, an inter mode is searched and then an intra mode and a SKIP block mode are searched. First, the inter mode performs a motion search on one 16x16 macroblock to a 16x16 block to obtain a coding cost for the Inter 16x16 mode (step 300), and then divides the 16x8 block into a 16x8 block to perform a motion search for the Inter 16x8 mode. The encoding cost is obtained (step 310). Then, the coding cost for the Inter 8x16 mode is obtained by performing a motion search by dividing the 16x16 macroblock into 8x16 blocks (step 320), and then, the motion cost is obtained by dividing the 16x16 macroblock into 8x8 blocks (step 330). . Specifically, in operation 330, a hatching cost for the P8x8 mode is obtained by performing a motion search by dividing the macro block into 8x8 blocks (step 332). The incubation for Inter 8x4 is performed by performing a motion search by dividing the macro block into 8x4 blocks. A step of obtaining a cost (step 334), a step of obtaining a hatching cost for the Inter 4x8 mode by performing a motion search by dividing into 4x8 blocks and a hatching cost for an Inter 4x4 mode by performing a motion search by dividing into a 4x4 block Step 338 is performed. As described above, if the motion search is performed with various variable block sizes, the image can be efficiently encoded according to the characteristics and the motion of the image. For example, if the motion of the image is small and the object is large, it is effective to estimate and encode the motion with a large block size when the motion of the image is large and the object is small.

Subsequently, when the motion search for the inter mode is completed, the motion search is performed in the intra mode, that is, the Intra 4x4, Intra 16x16, and SKIP block modes, to obtain the encoding cost for each (steps 340 to 360). Here, the SKIP block mode is a case in which the motion vector has a motion vector of origin (0,0) or Inter 16x16 mode, and the residual signal after DCT and quantization is all zero.

As described above, when the coding costs for all seven block modes are obtained, the block mode having the lowest cost among them is finally selected (step 370).

As described above, in the related art, a hatching cost is obtained by performing a motion search for a plurality of block modes, and among the macroblocks, the block mode in which the coding cost is determined to be the lowest, that is, the coding efficiency is determined to be the best. Decide on block mode. Therefore, a lot of gains can be seen in terms of bit rate, and image compression that provides better image quality at the same bit rate is possible.

 However, in the aspect of the image encoder, since the motion search is performed several times using a plurality of block modes to determine the block mode, the computational amount becomes very large, thereby increasing the complexity of the image encoder.

The technical problem to be solved by the present invention is to first determine whether to set the input macroblock to the SKIP block mode when determining the block mode for encoding by motion estimation in the image encoder, and then determine the macroblock determined to the SKIP block mode. For a high-speed block mode determination method that can greatly reduce the amount of computation by omitting the motion search for the inter mode.

Another object of the present invention is to provide a recording medium in which the fast block mode determination method is recorded by program code executable by a computer.

In order to achieve the above technical problem, in the fast block mode determination method according to the present invention in which a block mode for encoding is determined by motion estimation in an image encoder, an adaptive threshold value is determined in consideration of motion between neighboring frames. In step (b) and step (b) of determining whether to input the 16x16 macroblock into the SKIP block mode using the adaptive threshold value, if it is determined to be the SKIP block mode, encoding is performed through the motion search for the SKIP block mode. (C) obtaining a cost for each of the interblock modes through motion search for the interblock mode if the cost is not determined and the SKIP block mode is determined, and for each intrablock mode through the motion search for the intrablock mode. Comparing the coding costs obtained in steps (d) and (c) and (d) for obtaining the coding cost, And (e) determining the block mode having the depreciation cost as the final block mode for motion estimation of the 16x16 macroblock.

Hereinafter, a fast block mode determination method according to the present invention will be described with reference to the accompanying drawings.

Before describing the method of determining a block mode according to the present invention, the characteristics of the SKIP block mode will be described. Among the seven block modes, the SKIP block mode is used to effectively encode a portion repeated every frame like a background in a video. To determine 16x16 macroblock in SKIP block mode, the following four conditions must be met.

1. The motion compensation block must be 16x16.

2. The reference picture should be the previous picture.

3. The motion vector must be the same as the predicted motion vector.

4. The bit amount of surplus data must be zero.

On the other hand, it is very likely that the encoding cost for the SKIP block mode is lower than the encoding cost for the inter block modes Inter 16x16, Inter 8x16, Inter 16x8, and P8x8. In the fast block mode determination method according to the present invention, whether the SKIP block mode is first determined before performing the search in which the input macroblock is moved to the interblock mode with respect to the input macroblock, and then determined as the SKIP block mode. By omitting the motion search for the inter block mode, the image estimator can determine the block mode more quickly and simply in the image encoder.

4 is a flowchart illustrating an embodiment of a fast block mode determination method according to the present invention using the characteristics of the SKIP block mode as described above.

Referring to FIG. 4, the motion estimator (see FIG. 1) calculates an adaptive threshold Th _skip in consideration of motion between neighboring frames (operation 400).

The adaptive threshold Th _skip may be obtained using an average MAFD (Mean Absolute Frame difference) of pixel values of the difference image between the current image and the previous image. First, an average MAFD of pixel values of two neighboring images may be obtained as in Equation 1 below.

Where M and N represent the horizontal and vertical size of the image, X _{i, j} represent the (i, j) th pixel value in the current image, and Y _{i, j} represent the (i, j) th pixel in the previous image Indicates a value.

In general, MAFD contains a lot of information about the movement of consecutive frames. For example, if the MAFD has a large value, it indicates a motion boundary or a boundary of a moving object, while if the MAFD has a small value, it indicates that the object has a small sequence. Therefore, the MAFD will have a large value at the portion where the neighboring interframe movement occurs in the sequence. On the contrary, the MAFD will have a small value when there is little or no movement of the object between the two frames. By using the characteristics of the MAFD it is possible to set a threshold for determining the SKIP block mode. It is possible to extract more accurate SKIP block mode by adaptively applying the threshold set using MAFD for each extraction frame according to the degree of movement between neighboring frames in the sequence. The adaptive threshold value Th _skip using MAFD is obtained as shown in Equation 2 below.

Here, w denotes a weight component considering the minimum threshold (Th _skipmin), a threshold minimum value (Th _skipmin) is determined through measurement of the degree of the image quality deteriorates by the experiments of the SAD (0,0). As shown in Equation 2, if the adaptive threshold value (Th _skip ) obtained using _MAFD is larger than the threshold minimum value (Th _skipmin ), the value is set as the adaptive threshold value (Th _skip ), otherwise, the threshold minimum value (Th _{skipmin) is used.} ) Is set to the adaptive threshold Th _skip .

After operation 400, a motion search is performed on the input 16x16 macroblock in Inter 16x16 mode to obtain a motion vector and an encoding cost (operation 410). Here, the encoding cost is obtained through rate-distortion optimization considering the deterioration of image quality and the amount of generated bits as described above. Functions for measuring degradation include Sum of Absolute Difference (SAD), Sum of Absolute Transformed Difference (SATD), Sum of Squared Difference (SSD), Mean of Absolute Difference (MAD), or Lagrange Function. It is not limited to this. Among these, the coding cost measured using the SAD function is a value obtained by adding the absolute value of the difference between the predicted value and the actual pixel values of each pixel of the current subblock.

Subsequently, the Inter 16x16 mode motion vector and the encoding cost obtained in step 410 are the conditions of the SKIP block mode, that is, the motion vector is the origin (0,0) or the same as the predicted motion vector, and the encoding cost is the adaptive threshold ( It is determined whether or not less than Th _skip (step 420).

If the motion vector and the encoding cost of Inter 16x16 obtained in operation 410 satisfy the condition of the SKIP block mode in operation 420, the motion vector and the encoding cost are obtained by motion searching the currently input 16x16 macroblock in the SKIP block mode. Step 430).

On the other hand, if the motion vector and the encoding cost obtained in step 410 do not satisfy the conditions of the SKIP block mode, the motion vector and the encoding cost for each of the remaining inter modes, Ineter16x8 mode, Inter8x16 mode, and P8x8 mode, are sequentially searched. And first select the mode having the lowest coding cost (steps 440 to 470).

If the first selected mode in step 470 is the Inter 16x16 mode, it is checked whether the motion vector (MV) is the origin (0,0) (step 480), and if the motion vector is the origin (0,0), the first selected mode is selected. The mode is converted to the SKIP block mode (S430). As such, when the first selected mode is the Inter 16x16 mode, the operation may determine whether the motion vector MV is the origin (0,0) again, and again filter the non-filtered SKIP block mode in operation 420. In addition, the reason for determining whether the SKIP block mode is the Inter 16x16 mode as described above is that, as described above, the basic condition for determining the SKIP block mode is that the motion search block must be 16x16.

When the input 16x16 macroblock is determined as the primary selected block mode or SKIP block mode, the Intra 4x4 mode and the Intra 16x16 mode are sequentially searched to obtain respective motion vectors and encoding costs (steps 490 to 500).

The final block for the motion prediction of the 16x16 macroblock in which the encoding cost of the first selected block mode or the encoding cost of the SKIP block mode is compared with the encoding cost of the Intra 4x4 mode and the Intra 16x16 mode. Determine the mode.

As described above, the fast block mode determination method according to the present invention does not perform a calculation process for obtaining a motion search and encoding cost for the Ineter16x8 mode, the Inter8x16 mode, and the P8x8 mode for the macroblock determined as the SKIP block mode in step 420. Since it is not necessary, the computation amount of the image encoder can be greatly reduced. Particularly, the method for determining a fast block mode according to the present invention is used in a field such as a video, for example, a video call or a video conferencing system, in which there is not much motion and a sequence including a lot of flat areas, that is, a lot including a SKIP block mode. Excellent performance can be exhibited.

The invention can also be embodied as computer readable code on a computer readable recording medium. The computer-readable recording medium includes all kinds of recording devices in which data that can be read by a computer system is stored. Examples of computer-readable recording media include ROM, RAM, CD-ROM, magnetic tape, floppy disk, optical data storage, and the like, which are also implemented in the form of a carrier wave (for example, transmission over the Internet). It also includes. The computer readable recording medium can also be distributed over network coupled computer systems so that the computer readable code is stored and executed in a distributed fashion. And functional programs, codes and code segments for implementing the present invention can be easily inferred by programmers in the art to which the present invention belongs.

Although the preferred embodiments of the present invention have been shown and described above, the present invention is not limited to the above-described specific preferred embodiments, and the present invention belongs to the present invention without departing from the gist of the present invention as claimed in the claims. Various modifications can be made by those skilled in the art, and such changes are within the scope of the claims.

As described above, the fast block mode determination method according to the present invention first determines whether the SKIP block mode is prior to the motion search for the inter mode when determining the block mode for motion prediction, and intercepts the block determined as the SKIP block mode. Since the process of searching for motions can be omitted, the block mode can be determined quickly with less computation.

Claims (8)

A fast block mode determination method for determining a block mode for encoding through motion estimation in a video encoder, (a) obtaining an adaptive threshold in consideration of movement between neighboring frames; (b) using the adaptive threshold value, determining whether to input the 16 × 16 macroblock into the SKIP block mode; (c) If it is determined in SKIP block mode in step (b), the coding cost is obtained through motion search for the SKIP block mode, and if it is not determined as SKIP block mode, motion code for the inter modes is performed for each inter mode. Obtaining an encoding cost; (d) obtaining a coding cost for each of the intra modes through motion searching for the intra modes; And (e) comparing a coding cost obtained in steps (c) and (d) to determine a block mode having a minimum coding cost as a final block mode for motion prediction of the 16x16 macroblock. A fast block mode determination method. The method of claim 1, The adaptive threshold value is a fast block mode determination method, characterized in that the setting using the average MAFD (Mean Absolute Frame difference) of the pixel value for the difference between the current image and the previous image. 3. The method of claim 2, wherein the adaptive threshold is

Wherein Th _skip represents an adaptive threshold value and w ₁ represents a weight component in consideration of deterioration of image quality. 3. The method of claim 2, wherein the adaptive threshold is

Where Th _skip represents an adaptive threshold value, Th _skipmin represents a threshold minimum value, which is the minimum value of the adaptive threshold value considering the deterioration of image quality, and w ₂ represents a weight component considering Th _skipmin . Fast block mode determination method, characterized in that. The method of claim 4, wherein The threshold minimum value is determined by measuring the degree of deterioration of image quality by the experiment of SAD (0,0). The method of claim 1, wherein steps (b) and (c) (b1) obtaining a motion vector and an encoding cost by performing a motion search on the 16x16 macroblock in an inter 16x16 mode among inter modes; (b2) determining whether the motion vector obtained in step (b1) satisfies a condition of a predetermined SKIP block mode and whether the coding cost obtained in step (b1) is smaller than the adaptive threshold value; (b3) obtaining a coding cost by performing a motion search using the 16x16 macroblock in the SKIP block mode when the condition of the step (b2) is satisfied; And (b4) if the condition of step (b2) is not satisfied, performing a motion search using the 16x16 macroblock as an Inter 16x8 mode, an Inter8x16 mode, and a P8x8 mode, and obtaining a motion vector and an encoding cost for each mode. A fast block mode determination method, characterized in that. The method of claim 6, wherein the condition of the SKIP block mode And determining whether the motion vector obtained in the step (b1) is the origin (0,0) or the same as the predicted motion vector. The method of claim 6, (b5) if a motion vector and an encoding cost for each of the Inter 16x16 mode, the Inter 16x8 mode, the Inter8x16 mode, and the P8x8 mode are obtained, first selecting a block mode having the lowest encoding cost among them; (b6) checking whether the motion vector is the origin (0,0) when the first selected block mode is the Inter 16x16 mode; And and (b7) converting the first selected mode to the SKIP block mode if the motion vector is the origin (0,0) in step (b6).

Images