CN112437305B - Method and device for judging macro block type, image processing chip and electronic equipment - Google Patents

Abstract

The present invention provides a macroblock type determination method, device, image processing chip and electronic equipment. The determination method includes the following steps, for the current block: calculating the first determination flag, the second determination flag and the third determination flag of the skip mode, the first determination flag The cost, the second cost, and the third cost, as well as the intra-frame prediction cost, and the final jump motion vector of the skip mode; according to whether the first judgment flag, the second judgment flag, and the third judgment flag of the skip mode are of the skip type, the skip mode is the first Whether the temporary skip motion vector, the second temporary skip motion vector, and the sub-pixel motion vector after macroblock motion estimation are equal to the final skip motion vector of skip mode, and the difference between the first cost, the second cost, the third cost and the intra prediction cost is finally Determine the current macroblock type. The present invention can process in parallel and improve the accuracy of judging the type of skip macroblocks.

Description

Method, device, image processing chip and electronic device for determining macroblock type

technical field

The present invention relates to the technical field of video compression and coding, and in particular, to a method, device, image processing chip and electronic device for determining a macroblock type.

Background technique

H.264/AVC is a video compression standard jointly formulated by ISO/IEC and ITU-T, two international standardization organizations. It adopts many new compression technologies and has higher compression efficiency and better network adaptability. In the case of the same image quality, a smaller code stream is generated, thereby saving more bandwidth. Since H.264 adopts more coding techniques, the complexity of the algorithm has increased a lot, and the difficulty of hardware implementation has also increased.

An important feature of H.264 is the use of transform size mode selection and motion estimation. The current frame is predicted using the encoded reference frame, and each macroblock in the current frame is represented by a residual block and motion vector, which effectively eliminates the temporal correlation between video sequences. Among them, the H.264 standard introduces a SKIP block (skip macroblock) in the P frame encoding to effectively encode a large area of still areas and areas with consistent motion. The SKIP block is also a motion prediction compensation mode with a block size of 16x16. Especially in the case of a static scene or a constant moving background, the proportion of SKIP macroblocks is very large. To reduce the code rate, if it is known in advance that some macroblocks are SKIP macroblocks, it is not necessary to perform DCT transformation and quantization processing on the SKIP macroblocks, that is, it is not necessary to encode motion vectors and residuals, but only SKIP macroblocks are processed. Mark, when decoding, simply copy the corresponding macroblock from the reference frame according to the mark. From this, it can be seen that the judgment of the SKIP macroblock is very important to improve the coding effect and reduce the coding rate.

x264 is an open source H.264/AVC video encoding function library, which is one of the best video encoders to implement the H.264 standard. In the existing x264 coding model, there are three methods for judging the SKIP mode of the P frame. The following discussion assumes that the encoder only uses one forward reference frame, and the P frame uses only one division of 16x16.

Traditional scheme 1: Before the start of integer pixel motion estimation (ie IME), if one of the left neighbor block, the upper neighbor block, the upper left neighbor block, and the upper right neighbor block of the current macroblock is a SKIP block, the judgment of the SKIP block is performed. The function x264_macroblock_probe_pskip() (can be called the skip mode cost calculation method of skip macroblock), the specific operation is to obtain the reconstructed luminance macroblock and Chroma macroblock, then DCT transform, quantize, and zigzag scan the reconstructed luminance block and chrominance block in 4x4 units, and determine whether the current macroblock is a SKIP block according to the number of non-zero values after 4x4 block scanning .

Traditional scheme 2: After the current macroblock 16x16 mode sub-pixel motion estimation (ie FME) is completed, if the obtained mv is very close to the pskip_mv of the current macroblock, and the optimal cost value after motion estimation is less than a certain threshold value, Then execute the function x264_macroblock_probe_pskip() to determine whether the current macroblock is a skip block.

Traditional scheme 3: In the macroblock encoding process, after quantization, if the current macroblock type is P block and the partition size is 16x16, most of the quantized coefficients are 0, and the macroblock motion estimation mv is equal to the current macroblock. pskip_mv of the block, the current macroblock is determined as a skip block.

The traditional scheme 1 and the traditional scheme 2 judge that when calculating the pskip_mv of the skip mode, the type of the left macroblock needs to be dependent. This dependency will inevitably affect the parallel implementation of the hardware pipeline, which will lead to the extension of the hardware pipeline and reduce the real-time encoding of the encoder. speed. Although the traditional solution 3 can realize parallel processing to a certain extent, the premise that the judgment step of the traditional solution 3 becomes a SKIP block after the macroblock type is determined is that the current macroblock type is a P block. If the intra-frame prediction cost of the current macroblock is equal to the inter-frame prediction cost, and the final macroblock type is determined to be an I block, some SKIP blocks are easy to ignore. low, which makes the encoder unable to meet the application in the harsh environment of the code rate.

SUMMARY OF THE INVENTION

Based on the above situation, the main purpose of the present invention is to provide a macroblock type determination method, apparatus, image processing chip and electronic device, so as to improve the accuracy and parallel processing of skip macroblock type determination.

For achieving the above object, the technical scheme adopted in the present invention is as follows:

A method for determining a macroblock type to be encoded in a video frame, comprising the steps of: S100, using a motion vector prediction algorithm to calculate a temporary predicted motion vector of the current macroblock according to the motion vector of the non-left neighbor macroblock in front of the current macroblock; S200. Set a first temporary jump motion vector of the skip mode of the current macroblock, and use a skip mode cost calculation method to calculate the skip of the current macroblock according to the temporary predicted motion vector and the first temporary jump motion vector of the skip mode The first judgment mark and the first cost of the pattern; S300, temporarily set the type of the left neighbor macroblock of the current macroblock to be an inter-frame macroblock, and utilize the jump motion vector prediction algorithm according to the subpixel of the left neighbor macroblock The motion vector and the motion vector of the non-left neighbor macroblock are used to calculate a second temporary skip motion vector of the skip mode of the current macroblock, and the skip mode cost calculation method is used according to the temporary predicted motion vector and the skip mode. The second temporary skip motion vector calculates the second judgment flag and the second cost of the skip mode of the current macroblock; S400, perform sub-pixel motion estimation on the current macroblock to obtain the sub-pixel motion vector of the current macroblock, Use the skip mode cost calculation method to calculate the third judgment flag and the third cost of the skip mode of the current macroblock according to the temporary predicted motion vector and the sub-pixel motion vector of the current macroblock; S500, calculate the Intra prediction cost of the current macroblock; S600, after the type of the left neighbor macroblock is determined, use the skip motion vector prediction algorithm to calculate the motion vector according to the left neighbor macroblock and the non-left neighbor macroblock The final skip motion vector of the skip mode of the current macroblock; S700. If the first judgment flag of the skip mode is skip type, the first temporary skip motion vector of the skip mode is equal to the final skip motion vector of the skip mode, and the first cost is equal to the The difference between the intra-frame prediction costs is not greater than the first threshold, or, the second judgment flag of the skip mode is a skip type, the second temporary skip motion vector of the skip mode is equal to the final skip motion vector of the skip mode, and the second cost The difference from the intra-frame prediction cost is not greater than the first threshold, or the third judgment flag of the skip mode is the skip type, and the sub-pixel motion vector of the current macroblock is equal to the final skip motion vector of the skip mode And the difference between the third cost and the intra-frame prediction cost is not greater than the second threshold, then the current macroblock is determined to be a skipped macroblock.

Preferably, the step S400 further includes the following steps: obtaining the temporary inter-frame prediction cost of the current macroblock generated in the process of performing sub-pixel motion estimation on the current macroblock; the step S600 further includes the following steps: using The motion vector prediction algorithm calculates the final predicted motion vector of the current macroblock according to the motion vectors of the left neighbor macroblock and the non-left neighbor macroblock; uses the inter prediction block cost algorithm according to the final predicted motion vector Correcting the temporary inter-frame prediction cost with the difference between the temporary predicted motion vector to obtain the final inter-frame prediction cost of the current macroblock; in step S700, if the current macroblock is not determined to be a skip macroblock, then: Compare the size of the intra-frame prediction cost and the final inter-frame prediction cost, if the intra-frame prediction cost is less than the final inter-frame prediction cost, then determine the type of the current macroblock as an intra-frame macroblock, otherwise The type of the current macroblock is determined as an inter-frame macroblock.

Preferably, in the step S200, the first temporary skip motion vector of the skip mode of the current macroblock is set as a zero vector.

Preferably, the non-left neighbor macroblocks are the upper left neighbor macroblock, the upper neighbor macroblock and the upper right neighbor macroblock of the current macroblock, that is, three neighbor macros are taken for calculation.

Preferably, both the first threshold and the second threshold are 0.

Preferably, the partition mode of all macroblocks is 16x16 pixels.

The present invention also provides a device for determining the type of a macroblock to be encoded in a video frame, comprising: a first processing unit, configured to use a motion vector prediction algorithm to calculate the current macroblock according to the motion vector of the non-left neighbor macroblock before the current macroblock The temporary motion vector predictor of the macroblock; the second processing unit is used to set the first temporary skip motion vector of the skip mode of the current macroblock, and utilizes the skip mode cost calculation method according to the temporary motion vector predictor and the first temporary skip motion vector of the skip mode. A temporary skip motion vector is used to calculate the first judgment flag and the first cost of the skip mode of the current macroblock; a third processing unit is used to temporarily set the type of the left neighbor macroblock of the current macroblock as inter-frame macro block, using the skip motion vector prediction algorithm to calculate the second temporary skip motion vector of the skip mode of the current macroblock according to the sub-pixel motion vector of the left neighbor macroblock and the motion vector of the non-left neighbor macroblock, using The skip mode cost calculation method calculates the second judgment flag and the second cost of the skip mode of the current macroblock according to the temporary predicted motion vector and the second temporary skip motion vector of the skip mode; the fourth processing unit is used for Perform sub-pixel motion estimation on the current macroblock to obtain a sub-pixel motion vector of the current macroblock, and use the skip mode cost calculation method to calculate the temporary predicted motion vector and the sub-pixel motion vector of the current macroblock a third judgment flag and a third cost of the skip mode of the current macroblock; a fifth processing unit, used to calculate the intra prediction cost of the current macroblock; a sixth processing unit, used for the left neighbor macro After the block type is determined, the skip motion vector prediction algorithm is used to calculate the final skip motion vector of the skip mode of the current macroblock according to the motion vectors of the left neighbor macroblock and the non-left neighbor macroblock; the seventh processing unit, If the first judgment flag of the skip mode is skip type, the first temporary skip motion vector of skip mode is equal to the final skip motion vector of skip mode, and the difference between the first cost and the intra prediction cost is not greater than the first The threshold, or the second judgment flag of the skip mode is the skip type, the second temporary skip motion vector of the skip mode is equal to the final skip motion vector of the skip mode, and the difference between the second cost and the intra prediction cost is not greater than all The first threshold, or the third judgment flag of the skip mode is the skip type, the sub-pixel motion vector of the current macroblock is equal to the final skip motion vector of the skip mode, and the third cost and the intra prediction cost are combined If the difference is not greater than the second threshold, the current macroblock is determined as a skipped macroblock.

Preferably, the fourth processing unit is further configured to: obtain the temporary inter-frame prediction cost of the current macroblock generated in the process of performing sub-pixel motion estimation on the current macroblock; the sixth processing unit is further configured to: : use the motion vector prediction algorithm to calculate the final predicted motion vector of the current macroblock according to the motion vectors of the left neighbor macroblock and non-left neighbor macroblock; use the inter prediction block cost algorithm to calculate the final predicted motion vector according to the final prediction The difference between the motion vector and the temporary predicted motion vector corrects the temporary inter-frame prediction cost to obtain the final inter-frame prediction cost of the current macroblock; the seventh processing unit is also used for: if the current macroblock has not been determined For skipping macroblocks, compare the size of the intra-frame prediction cost and the final inter-frame prediction cost, and if the intra-frame prediction cost is less than the final inter-frame prediction cost, determine the type of the current macroblock as frame Intra macroblock, otherwise the type of the current macroblock is determined as an inter macroblock.

Preferably, the second processing unit is further configured to: set the first temporary skip motion vector of the skip mode of the current macroblock to a zero vector.

Preferably, the non-left neighbor macroblocks are the upper left neighbor macroblock, the upper neighbor macroblock and the upper right neighbor macroblock before the current macroblock, that is, three neighbor macros are taken for calculation.

Preferably, the partition mode of all macroblocks is 16x16 pixels.

The present invention also provides an image processing chip, which is used for executing any one of the macroblock type determination methods.

The invention also provides an image processing electronic device, including the image processing chip.

【Beneficial effects】

The present invention pre-judges the current macroblock type through the first judgment flag and the first cost of the skip mode, the second judgment flag and the second cost of the skip mode, and the third judgment flag and the third cost of the skip mode, and then according to The above parameters and the final skip motion vector of skip mode, the intra prediction cost, the first threshold and the second threshold and other parameters finally determine the macroblock type, which solves the problem of skipping macroblocks when the intra prediction cost and the inter prediction cost are equivalent in the prior art. The problem that the block is misjudged as an intra-frame prediction macroblock greatly satisfies the application of the encoder in the environment with strict code rate requirements. In addition, the present invention can also process macroblock coding and jump macroblock type judgment in parallel, shorten the hardware pipeline, save certain hardware design logic, and improve the real-time coding speed of the encoder.

Other beneficial effects of the present invention will be illustrated in the specific embodiments through the introduction of specific technical features and technical solutions. Those skilled in the art should be able to understand the technical features and technical solutions through the introduction of these technical features and technical solutions. beneficial technical effects.

Description of drawings

Preferred embodiments according to the present invention will be described below with reference to the accompanying drawings. In the picture:

FIG. 1 is a schematic diagram of an embodiment of a hardware encoder that executes a method for determining a macroblock type according to the present invention.

FIG. 2 is a schematic diagram of parallel processing of multiple macroblocks

3 is a partial flowchart of a method for determining a macroblock type according to an embodiment of the present invention

4 is a partial flowchart of a method for determining a macroblock type according to an embodiment of the present invention

Figure 5 is a schematic diagram of the current macroblock in the current frame and its preceding neighbor macroblocks (the left macroblock is unavailable)

FIG. 6 is a schematic diagram of the current macroblock in the current frame and its preceding neighbor macroblocks (the left macroblock is available)

Detailed ways

1 is a schematic diagram of an embodiment of a hardware encoder for implementing the method for determining the macroblock type of the present invention. The encoder includes an integer-pixel motion estimation (IME) unit, a sub-pixel motion estimation (FME) unit, intra-frame prediction and transformation. A quantization unit and a macroblock type judgment unit, wherein the integer pixel motion estimation unit needs to execute the unique macroblock type of the present invention in addition to executing the integer pixel motion estimation algorithm common in the prior art to calculate the integer pixel motion vector of the macroblock Pre-judgment step 1; the sub-pixel motion estimation unit needs to execute the unique macro of the present invention in addition to executing the sub-pixel motion estimation algorithm common in the prior art to calculate the sub-pixel motion vector of the macroblock and the corresponding inter-frame prediction cost. Block type pre-judgment step 2 and pre-judgment step 3; the intra-frame prediction, transform and quantization unit executes the existing common intra-frame prediction cost algorithm to calculate the intra-frame prediction cost of the macroblock; the macroblock type judgment unit is used in the macroblock type prediction. Based on the judgment step 1, the prejudgment step 2 and the prejudgment step 3, the final judgment step of the macroblock type is further performed to finally judge the type of the macroblock, such as skip macroblock, intra macroblock or inter macroblock type. The above macroblock type pre-judgment step 1, pre-judgment step 2 and pre-judgment step 3 will be described in detail below.

As shown in Figure 2, in order to fully realize the parallel execution of macroblock type determination, from the perspective of a macroblock, the macroblock will be sequentially processed by the whole-pixel motion estimation unit, the sub-pixel motion estimation unit, the intra-frame prediction and the transform and quantization unit. After processing with the macroblock type judgment unit (as shown in Figure 2, the above processing process is abbreviated as IME, FME, Intra prediction, MC+TQ, mb_type determination) to obtain the final macroblock type, for example, macroblock 1 is After the whole pixel motion estimation unit is processed, it is sent to the sub-pixel motion estimation unit for processing until the macroblock type of macroblock 1 is obtained after being processed by the macroblock type judgment unit; and from the perspective of each unit of the hardware encoder, It always processes a macroblock before processing the next macroblock. For example, after processing macroblock 1, the whole pixel motion estimation unit processes the adjacent next macroblock 2 until all pending macroblocks are processed. macroblock. In this way, each unit of the hardware encoder can process different macroblocks in parallel.

3 and 4 illustrate a method for determining the type of a macroblock to be encoded in a video frame according to an embodiment of the present invention, which includes the following steps.

The macroblock type pre-judgment step 1 includes step S1 and step S2, and is performed by an integer-pixel motion estimation unit. In one embodiment, the video object processed by this judgment method is a video sequence factory_1080p[1920x1080].yuv, and the encoding is set to: IDR=32 (IDR represents the interval between two I frames), QP=30 (QP is quantization coefficient value), the current macroblock may be, for example, the 20th macroblock of the 35th macroblock row of the first P frame of the video sequence.

S100. Use a motion vector prediction algorithm to calculate a temporary predicted motion vector mvp_tmp of the current macroblock according to the motion vector of the non-left neighbor macroblock in front of the current macroblock.

Specifically, as shown in FIG. 5 is a schematic diagram of the current macroblock in the current frame and the neighboring macroblocks in front of it. The neighboring macroblocks in front of the current macroblock include left-neighboring macroblocks and non-left-neighboring macroblocks. The left neighbor macroblock includes the upper left neighbor macroblock, the upper neighbor macroblock and the upper right neighbor macroblock of the current macroblock.

When the integer pixel motion estimation unit processes the current macroblock, the macroblock type of the left neighbor macroblock has not been calculated, and the sub-pixel motion vector of the left neighbor macroblock has not been calculated. Therefore, the left neighbor macroblock cannot be used. Temporary prediction motion vector mvp_tmp for predicting the current macroblock. In this step, the motion vector prediction algorithm is used to obtain the motion vector (referring to the final motion vector, calculated by the macroblock type judgment unit according to the motion vector (referring to the final motion vector) of the upper left neighbor macroblock, the upper neighbor macroblock and the upper right neighbor macroblock in front of the current macroblock ) Calculate the temporary predicted motion vector mvp_tmp of the current macroblock (for example, in one embodiment, calculate to obtain mvp_tmp=[0,0]), and the temporary predicted motion vector mvp_tmp will be used to determine the position of the search window. A matching block is searched within this search window to compute the associated motion vector or cost. The motion vector prediction algorithm may be a prediction algorithm specified in the H.264 standard, and is used to predict the motion vector of the current macroblock according to the motion vectors of the neighboring macroblocks.

S200. Set the first temporary skip motion vector pskip_mv_tmp0 of the skip mode of the current macroblock, and use the skip mode cost calculation method to calculate the first temporary skip motion vector pskip_mv_tmp0 of the skip mode of the current macroblock according to the temporary predicted motion vector mvp_tmp and the first temporary skip motion vector pskip_mv_tmp0 of the skip mode. A judgment flag pre_pskip_flag0 and the first cost pre_pskip_cost0. The first judgment flag pre_pskip_flag0 of skip mode includes two types: skip macroblock type and non-skip macroblock type. For example, when the first judgment flag pre_pskip_flag0 of skip mode is 1, it represents the skip macroblock type, and when it is 0, it represents non-skip macroblock type. Macroblock type.

Specifically, the first temporary skip motion vector pskip_mv_tmp0 of skip mode is set to zero vector (0,0) for common static scenes, as the first candidate skip motion vector, and the skip mode cost calculation method function in the original x264 coding model is used. According to the temporary predicted motion vector mvp_tmp and the first temporary skip motion vector pskip_mv_tmp0 of the skip mode, the first judgment flag pre_pskip_flag0 and the first cost pre_pskip_cost0 of the skip mode are calculated, wherein pre_pskip_flag0=x264_macroblock_probe_pskip() (parameters include mvp_tmp and pskip_mv_tmp0), in The first cost pre_pskip_cost0 is also obtained during the execution of the x264_macroblock_probe_pskip() function. The specific calculation process of the first cost pre_pskip_cost0 is as follows: the position of the search window is determined by the temporary predicted motion vector mvp_tmp, then the position of the motion compensation block is determined in the search window according to the first temporary skip motion vector pskip_mv_tmp0 of the skip mode, and then the calculation The SAD (Sum of Absolute Error) between this motion compensation block and the current macroblock. Since the cost of skipping macroblocks includes: the coding cost of SAD, the first cost pre_pskip_cost0 can be obtained after determining the SAD of skipping macroblocks. For example, in one embodiment, the following calculation results are obtained: pre_pskip_flag0=1, pre_pskip_cost0=1292.

The macroblock type pre-judgment step 2 includes step S300 and is performed by the sub-pixel motion estimation unit.

S300. Temporarily set (ie temporarily assume) the type of the left neighbor macroblock of the current macroblock to be an inter-frame macroblock (ie, a P block), and use the skip motion vector prediction algorithm according to the sub-pixel motion vector qmv_L of the left neighbor macroblock, and The motion vector of the non-left neighbor macroblock (ie, the final motion vector) calculates the second temporary skip motion vector pskip_mv_tmp1 of the skip mode of the current macroblock, and uses the skip mode cost calculation method according to the temporary predicted motion vector mvp_tmp and the second temporary skip of the skip mode The motion vector pskip_mv_tmp1 calculates the second judgment flag pre_pskip_flag1 and the second cost pre_pskip_cost1 of the skip mode of the current macroblock. Among them, the second judgment flag pre_pskip_flag1 of skip mode includes two types: skip macroblock type and non-skip macroblock type. For example, when the second judgment flag pre_pskip_flag1 of skip mode is 1, it represents the skip macroblock type, and when it is 0, it represents non-skip macroblock type. Macroblock type. For example, in one embodiment, the following calculation results are obtained: pskip_mv_tmp1=[0,1], pre_pskip_flag1=1, pre_pskip_cost1=1302.

Specifically, as shown in FIG. 6 , at this time, the left-neighbor macroblock has been processed by the sub-pixel motion estimation unit, that is, the sub-pixel motion vector qmv_L of the left-neighbor macroblock is obtained. Temporarily set the type of the left neighbor macroblock to be an inter-frame macroblock (ie, P block), and use the skip motion vector prediction algorithm according to the sub-pixel motion vector qmv_L of the left neighbor macroblock, and the upper left neighbor macroblock, upper neighbor macroblock and The motion vector of the upper right neighbor macroblock calculates the second skip motion vector pskip_mv_tmp1 of the current macroblock as the second candidate skip motion vector. The skip motion vector prediction algorithm may be a prediction algorithm specified in the H.264 standard, which is used to predict the skip motion vector of the current macroblock according to the motion vector of the neighbor macroblock, so as to calculate the type identifier and cost of the current macroblock, for example, the skip motion vector The prediction algorithm is specifically: the jump motion vector prediction algorithm is specifically: the sub-pixel motion vector qmv_L of the left neighbor macroblock, and the motion vectors of the upper left neighbor macroblock and the upper neighbor macroblock are sorted in descending order, and then select The motion vector at the intermediate position is used as the second skip motion vector pskip_mv_tmp1. Then, use the skip mode cost calculation method to calculate the second judgment flag pre_pskip_flag1 and the second cost pre_pskip_cost1 of the skip mode of the current macroblock according to the temporary predicted motion vector mvp_tmp and the second temporary skip motion vector pskip_mv_tmp1 of the skip mode, for example, use the original x264 encoding The skip mode cost calculation function in the model calculates the second judgment flag pre_pskip_flag1 and the second cost pre_pskip_cost1 of the skip mode according to the temporary predicted motion vector mvp_tmp and the second temporary skip motion vector pskip_mv_tmp1 of the skip mode, wherein, pre_pskip_flag1=x264_macroblock_probe_pskip()( The parameters include mvp_tmp and pskip_mv_tmp1), and the second cost pre_pskip_cost1 is also obtained during the execution of the x264_macroblock_probe_pskip() function. Among them, the specific calculation process of the second cost pre_pskip_cost1 is as follows: the position of the search window is determined by the temporary predicted motion vector mvp_tmp, and then the position of the motion compensation block is determined in the search window according to the second temporary skip motion vector pskip_mv_tmp1 of the skip mode, and then calculated The SAD (Sum of Absolute Error) between this motion compensation block and the current macroblock. Since the cost of skipping macroblocks includes the coding cost of SAD, the second cost pre_pskip_cost1 can be obtained after determining the SAD of skipping macroblocks.

The macroblock type pre-judgment step 3 includes step S400 and is performed by the sub-pixel motion estimation unit.

S400, perform subpixel motion estimation on the current macroblock to obtain the subpixel motion vector qmv of the current macroblock, and calculate the jump of the current macroblock according to the temporary predicted motion vector mvp_tmp and the subpixel motion vector qmv of the current macroblock using the skip mode cost calculation method The third judgment flag of the mode is common_pskip_flag and the third cost is common_pskip_cost.

Specifically, the sub-pixel motion estimation unit performs common sub-pixel motion estimation on the current macroblock, and can obtain the sub-pixel motion vector qmv of the current macroblock and the corresponding optimal cost of inter-frame prediction (may be referred to as inter-frame cost for short) i_cost_inter1, as a temporary inter-frame cost. Next, use the skip mode cost calculation function in the original x264 coding model to calculate the third judgment flag common_pskip_flag and the third cost common_pskip_cost of skip mode according to the temporary predicted motion vector mvp_tmp and the sub-pixel motion vector qmv, where common_pskip_flag=x264_macroblock_probe_pskip() (parameters include mvp_tmp and qmv), get common_pskip_cost during the execution of the x264_macroblock_probe_pskip() function.

In addition, calculate the optimal cost of inter-frame prediction i_cost_inter1, wherein the specific calculation process of the optimal cost of inter-frame prediction i_cost_inter1 is as follows: determine the position of the search window through the temporary prediction motion vector mvp_tmp, and then use the sub-pixel motion vector in the search window according to the sub-pixel motion vector. qmv determines the position of the motion compensation block and then calculates the SAD (sum of absolute error) between the motion compensation block and the current macroblock. Because, the optimal cost of inter prediction i_cost_inter1 contains two parts: one part is the coding cost of the motion vector of the macroblock (the difference between the sub-pixel motion vector qmv and the temporary prediction motion vector mvp_tmp) (for example, according to the sub-pixel motion vector qmv and the temporary prediction The difference between the motion vectors mvp_tmp is obtained by looking up the table to obtain the corresponding cost), part of which is the coding cost of SAD. Therefore, after determining the motion vector cost and SAD of the macroblock, the optimal cost i_cost_inter1 for temporary inter-frame prediction can be calculated. For example, in one embodiment, the following calculation results are obtained: qmv=[0,-1], i_cost_inter1=1290, common_pskip_flag=1, common_pskip_cost=1266.

S500. Calculate the intra prediction cost of the current macroblock.

Perform common intra-frame prediction (Intra) on the current macroblock, then dct transform and quantize, and calculate the optimal cost value i_cost_intra for intra-frame prediction. This step S500 may adopt the method specified by the H.264 standard. For example, in one embodiment, the following calculation result is obtained: i_cost_intra=1336.

The final determination step of the macroblock type, including steps S600 to S900, is performed by the macroblock type determination unit.

S600, after the type of the left neighbor macroblock is determined, use the motion vector prediction algorithm to calculate the final predicted motion vector mvp_real of the current macroblock according to the motion vectors of the left neighbor macroblock and the non-left neighbor macroblock, and utilize the skip motion vector prediction algorithm according to The motion vectors of the left neighbor macroblock and the non-left neighbor macroblock calculate the final skip motion vector pskip_mv_real of the skip mode of the current macroblock.

After processing the left neighbor macroblock, the macroblock type determination unit obtains the final macroblock type and final motion vector of the left neighbor macroblock, and then the macroblock type determination unit processes the current macroblock. Use the motion vector prediction algorithm to calculate the final predicted motion vector mvp_real of the current macroblock according to the final motion vectors of the left neighbor macroblock, the upper left neighbor macroblock, the upper neighbor macroblock and the upper right neighbor macroblock, and use the skip motion vector prediction algorithm to calculate the final predicted motion vector mvp_real of the current macroblock according to the left The final motion vectors of the neighbor macroblock, the upper left neighbor macroblock, the upper neighbor macroblock, and the upper right neighbor macroblock calculate the final skip motion vector pskip_mv_real for the skip mode of the current macroblock. For example, in one embodiment, the following calculation results are obtained: mvp_real=[0,0], pskip_mv_real=[0,0].

S700. Use an inter-frame prediction block cost algorithm to correct the temporary inter-frame prediction cost i_cost_inter1 according to the difference between the final predicted motion vector mvp_real and the temporary predicted motion vector mvp_tmp, to obtain the final inter-frame prediction cost i_cost_inter of the current macroblock.

Specifically, as can be seen from the previous introduction, the final inter prediction cost i_cost_inter of a macroblock includes two parts, the first part is the encoding cost of the macroblock’s motion vector (the difference between the sub-pixel motion vector qmv and the final predicted motion vector mvp_real) ( For example, according to the difference between the sub-pixel motion vector qmv and the temporary prediction motion vector mvp_tmp, the corresponding cost is obtained by looking up the table), the second part is the coding cost of SAD (same as SAD in step S4), therefore, it can be used in the temporary frame The final inter-frame prediction cost i_cost_inter is obtained by updating the first part of the coding cost based on the optimal inter-prediction cost i_cost_inter1, where i_cost_inter=i_cost_inter1+(mvp_real-mvp_tmp) coding cost. For example, in one embodiment, the following calculation result is obtained: mvp_real=mvp_tmp, so i_cost_inter is equal to i_cost_inter1.

S800. If the first judgment flag pre_pskip_flag0 of the skip mode is the skip type (ie pre_pskip_flag0=1), the first temporary skip motion vector pskip_mv_tmp0 of the skip mode is equal to the final skip motion vector pskip_mv_real of the skip mode (ie pskip_mv_tmp0=pskip_mv_real) and the first cost The difference between pre_pskip_cost0 and the intra prediction cost i_cost_intra is not greater than the first threshold PRE_PSKIP_TH (ie (pre_pskip_cost0-i_cost_intra)<=PRE_PSKIP_TH), or,

The second judgment flag pre_pskip_flag1 of the skip mode is the skip type (ie pre_pskip_flag1=1), the second temporary skip motion vector pskip_mv_tmp1 of the skip mode is equal to the final skip motion vector pskip_mv_real of the skip mode (ie pskip_mv_tmp1=pskip_mv_real), and the second cost pre_pskip_cost1 and frame The difference between the intra-prediction costs i_cost_intra is not greater than the first threshold PRE_PSKIP_TH (ie (pre_pskip_cost1-i_cost_intra)<=PRE_PSKIP_TH), or,

The third judgment flag common_pskip_flag of the skip mode is the skip type (ie common_pskip_flag=1), the sub-pixel vector qmv of the current macroblock is equal to the final skip motion vector pskip_mv_real (ie qmv=pskip_mv_real) of the skip mode, and the third cost common_pskip_cost and intra prediction The difference between the costs i_cost_intra is not greater than the second threshold COMMON_PSKIP_TH (ie (common_pskip_cost-i_cost_intra)<=COMMON_PSKIP_TH),

Then the current macroblock is determined as a skip macroblock, otherwise it goes to step S900 for further determination. Among them, PRE_PSKIP_TH=0, COMMON_PSKIP_TH=0. Wherein, the first threshold PRE_PSKIP_TH and the second threshold COMMON_PSKIP_TH can be any value within the range of values (-2 ⁹ to 2 ⁹ -1) that can be represented by 10-bit signed numbers, and can be set to suitable value.

In the foregoing embodiment, since pre_pskip_flag0=1, pskip_mv_tmp0=pskip_mv_real, and (pre_pskip_cost-i_cost_intra)<PRE_PSKIP_TH, it is determined that the current macroblock is a skip macroblock.

In one embodiment, the type of the skip macroblock may be determined according to the first judgment flag pre_pskip_flag0 of the skip mode, the first temporary skip motion vector pskip_mv_tmp0 of the skip mode, and the first cost pre_pskip_cost0. If it is not determined that the current macroblock type is skip As a result of the macroblock, continue to judge the type of the skip macroblock according to the second judgment flag pre_pskip_flag1 of the skip mode, the second temporary skip motion vector pskip_mv_tmp1 of the skip mode, and the second cost pre_pskip_cost1, if the current macroblock type is still not obtained. As a result of skipping the macroblock, the type of skipping macroblock is judged according to the third judgment flag common_pskip_flag of the skip mode, the sub-pixel vector qmv of the current macroblock, and the third cost common_pskip_cost.

S900: Determine the intra prediction cost i_cost_intra of the current macroblock and the final inter prediction cost i_cost_inter, if i_cost_intra<i_cost_inter, then determine the current macroblock as an intra prediction block, otherwise determine as an inter prediction block.

The present invention also provides a device for determining the type of macroblocks to be encoded in a video frame, including:

a first processing unit, used for calculating the temporary predicted motion vector of the current macroblock according to the motion vector of the non-left neighbor macroblock in front of the current macroblock using a motion vector prediction algorithm;

The second processing unit is configured to set the first temporary skip motion vector of the skip mode of the current macroblock, and use the skip mode cost calculation method to calculate the temporary predicted motion vector and the first temporary skip motion vector of the skip mode The first judgment flag and the first cost of the skip mode of the current macroblock;

a third processing unit, configured to temporarily set the type of the left neighbor macroblock of the current macroblock to be an inter macroblock, and use a skip motion vector prediction algorithm according to the sub-pixel motion vector of the left neighbor macroblock and the The motion vector of the non-left neighbor macroblock calculates the second temporary skip motion vector of the skip mode of the current macroblock, and uses the skip mode cost calculation method according to the temporary predicted motion vector and the second temporary skip motion vector of the skip mode Calculate the second judgment flag and the second cost of the skip mode of the current macroblock;

a fourth processing unit, configured to perform sub-pixel motion estimation on the current macroblock to obtain a sub-pixel motion vector of the current macroblock, and use the skip mode cost calculation method according to the temporary predicted motion vector and the current macroblock The sub-pixel motion vector of the block calculates the third judgment flag and the third cost of the skip mode of the current macroblock;

a fifth processing unit, configured to calculate the intra prediction cost of the current macroblock;

The sixth processing unit is configured to use the skip motion vector prediction algorithm to calculate the current macroblock according to the motion vectors of the left neighbor macroblock and the non-left neighbor macroblock after the type of the left neighbor macroblock is determined. the final jump motion vector for the jump mode;

A seventh processing unit, used for if the first judgment flag of the skip mode is a skip type, the first temporary skip motion vector of the skip mode is equal to the final skip motion vector of the skip mode, and the first cost and the intra prediction cost are combined difference is not greater than the first threshold, or,

The second judgment flag of the skip mode is the skip type, the second temporary skip motion vector of the skip mode is equal to the final skip motion vector of the skip mode, and the difference between the second cost and the intra prediction cost is not greater than the first threshold ,or,

The third judgment flag of the skip mode is the skip type, the sub-pixel motion vector of the current macroblock is equal to the final skip motion vector of the skip mode, and the difference between the third cost and the intra prediction cost is not greater than the second threshold,

Then, the current macroblock is determined as a skip macroblock.

It should be noted that, in the present invention, step numbers (letters or numbers) are used to refer to some specific method steps, which are only for the purpose of convenience and brevity of description, and are not intended to limit these method steps with letters or numbers. Order. Those skilled in the art can understand that the sequence of related method steps should be determined by the technology itself, and should not be unduly limited due to the existence of step numbers.

Those skilled in the art can understand that, under the premise of no conflict, the above preferred solutions can be freely combined and superimposed.

It should be understood that the above-mentioned embodiments are only exemplary rather than restrictive, and those skilled in the art can make various obvious or equivalent to the above-mentioned details without departing from the basic principles of the present invention. Modifications or substitutions will be included within the scope of the claims of the present invention.

Claims (11)

1. a method for determining a macroblock type to be encoded in a video frame, comprising the steps of: S100, utilize the motion vector prediction algorithm to calculate the temporary predicted motion vector of the current macroblock according to the motion vector of the non-left neighbor macroblock in front of the current macroblock; S200. Set a first temporary jump motion vector of the skip mode of the current macroblock, and use a skip mode cost calculation method to calculate the skip of the current macroblock according to the temporary predicted motion vector and the first temporary jump motion vector of the skip mode The first judgment flag and the first cost of the mode; S300. Assuming that the type of the left neighbor macroblock of the current macroblock is an inter-frame macroblock, use a skip motion vector prediction algorithm according to the sub-pixel motion vector of the left neighbor macroblock and the motion of the non-left neighbor macroblock vector to calculate the second temporary skip motion vector of the skip mode of the current macroblock, and use the skip mode cost calculation method to calculate the current macroblock according to the temporary predicted motion vector and the second temporary skip motion vector of the skip mode. the second judgment flag and the second cost of the skip mode; S400. Perform sub-pixel motion estimation on the current macroblock to obtain a sub-pixel motion vector of the current macroblock, and use the skip mode cost calculation method according to the temporary predicted motion vector and the sub-pixel motion of the current macroblock Vector calculation of the third judgment flag and the third cost of the skip mode of the current macroblock; S500, calculating the intra prediction cost of the current macroblock; S600. After the type of the left-neighbor macroblock is determined, use the skip motion vector prediction algorithm to calculate the final skip of the skip mode of the current macroblock according to the motion vectors of the left-neighbor macroblock and the non-left-neighbor macroblock motion vector; S700. If the first judgment flag of the skip mode is the skip type, the first temporary skip motion vector of the skip mode is equal to the final skip motion vector of the skip mode, and the difference between the first cost and the intra prediction cost is not greater than the first threshold, or, The second judgment flag of the skip mode is the skip type, the second temporary skip motion vector of the skip mode is equal to the final skip motion vector of the skip mode, and the difference between the second cost and the intra prediction cost is not greater than the first threshold ,or, The third judgment flag of the skip mode is the skip type, the sub-pixel motion vector of the current macroblock is equal to the final skip motion vector of the skip mode, and the difference between the third cost and the intra prediction cost is not greater than the second threshold, Then, the current macroblock is determined as a skip macroblock. 2. The method for determining a macroblock type according to claim 1, wherein, Step S400 also includes the following steps: obtaining the temporary inter prediction cost of the current macroblock generated in the process of performing sub-pixel motion estimation on the current macroblock; Step S600 further includes the following steps: Using the motion vector prediction algorithm to calculate the final predicted motion vector of the current macroblock according to the motion vectors of the left neighbor macroblock and the non-left neighbor macroblock; Correct the temporary inter prediction cost according to the difference between the final predicted motion vector and the temporary predicted motion vector by using an inter prediction block cost algorithm to obtain the final inter prediction cost of the current macroblock; In step S700, if the current macroblock is not determined to be a skipping macroblock, compare the size of the intra-frame prediction cost and the final inter-frame prediction cost, if the intra-frame prediction cost is less than the final inter-frame prediction cost, Then, the type of the current macroblock is determined as an intra macroblock, otherwise, the type of the current macroblock is determined as an inter macroblock. 3. The determination method according to claim 1, characterized in that, In step S200, the first temporary skip motion vector of the skip mode of the current macroblock is set to a zero vector. 4. The determination method according to claim 1, wherein, The non-left neighbor macroblocks are the upper left neighbor macroblock, the upper neighbor macroblock and the upper right neighbor macroblock of the current macroblock. 5. The determination method according to any one of claims 1-4, characterized in that, The first threshold and the second threshold are both 0. 6. The determination method according to any one of claims 1-4, wherein, The partition mode for all macroblocks is 16

16 pixels. 7. A device for determining the type of macroblocks to be encoded in a video frame, comprising: a first processing unit, used for calculating the temporary predicted motion vector of the current macroblock according to the motion vector of the non-left neighbor macroblock in front of the current macroblock using a motion vector prediction algorithm; The second processing unit is configured to set the first temporary skip motion vector of the skip mode of the current macroblock, and use the skip mode cost calculation method to calculate the temporary motion vector predictor and the first temporary skip motion vector of the skip mode The first judgment flag and the first cost of the skip mode of the current macroblock; The third processing unit is configured to assume that the type of the left neighbor macroblock of the current macroblock is an inter macroblock, and use a skip motion vector prediction algorithm according to the sub-pixel motion vector of the left neighbor macroblock and the non-left The second temporary skip motion vector of the skip mode of the current macroblock is calculated from the motion vector of the neighbor macroblock, and the calculated skip motion vector is calculated according to the temporary predicted motion vector and the second temporary skip motion vector of the skip mode by using the skip mode cost calculation method. Describe the second judgment flag and the second cost of the skip mode of the current macroblock; a fourth processing unit, configured to perform sub-pixel motion estimation on the current macroblock to obtain a sub-pixel motion vector of the current macroblock, and use the skip mode cost calculation method according to the temporary predicted motion vector and the current macroblock The sub-pixel motion vector of the block calculates the third judgment flag and the third cost of the skip mode of the current macroblock; a fifth processing unit, configured to calculate the intra prediction cost of the current macroblock; The sixth processing unit is configured to use the skip motion vector prediction algorithm to calculate the current macroblock according to the motion vectors of the left neighbor macroblock and the non-left neighbor macroblock after the type of the left neighbor macroblock is determined. the final jump motion vector for the jump mode; A seventh processing unit, used for if the first judgment flag of the skip mode is a skip type, the first temporary skip motion vector of the skip mode is equal to the final skip motion vector of the skip mode, and the first cost and the intra prediction cost are combined difference is not greater than the first threshold, or, The second judgment flag of the skip mode is the skip type, the second temporary skip motion vector of the skip mode is equal to the final skip motion vector of the skip mode, and the difference between the second cost and the intra prediction cost is not greater than the first threshold ,or, The third judgment flag of the skip mode is the skip type, the sub-pixel motion vector of the current macroblock is equal to the final skip motion vector of the skip mode, and the difference between the third cost and the intra prediction cost is not greater than the second threshold, Then, the current macroblock is determined as a skip macroblock. 8. The determination device according to claim 7, characterized in that, The fourth processing unit is also used for: obtaining the temporary inter prediction cost of the current macroblock generated in the process of performing sub-pixel motion estimation on the current macroblock; The sixth processing unit is also used for: Using the motion vector prediction algorithm to calculate the final predicted motion vector of the current macroblock according to the motion vectors of the left neighbor macroblock and the non-left neighbor macroblock; Correct the temporary inter prediction cost according to the difference between the final predicted motion vector and the temporary predicted motion vector by using an inter prediction block cost algorithm to obtain the final inter prediction cost of the current macroblock; The seventh processing unit is also used for: If the current macroblock is not determined to be a skip macroblock, compare the size of the intra-frame prediction cost and the final inter-frame prediction cost, and if the intra-frame prediction cost is less than the final inter-frame prediction cost, the current The type of the macroblock is determined as an intra macroblock, otherwise the type of the current macroblock is determined as an inter macroblock. 9. The determination device according to claim 7, characterized in that, The second processing unit is further configured to: set the first temporary skip motion vector of the skip mode of the current macroblock to a zero vector. 10. The determination device according to claim 7, characterized in that, The non-left neighbor macroblocks are the upper left neighbor macroblock, the upper neighbor macroblock and the upper right neighbor macroblock of the current macroblock. 11. The determination device according to any one of claims 7-10, wherein, The partition mode for all macroblocks is 16

16 pixels.

Images