KR20090041944A - Method and apparatus for motion estimation using mode information of neighbor blocks - Google Patents

Abstract

A method and an apparatus for motion estimation using mode information of neighbor blocks are provided to design a low electrical power and low cost moving picture encoder by reducing an amount of calculation for motion estimation. A mode information statistics table(110) stores information statistically organized from results of experiments by using various conditions for various video images, for example, a resolution, a frame rate, and a Qp value, etc. A motion estimation unit(120) obtains mode information of a neighboring block for a motion estimation about the current block. By using the mode information and mode information statistics table of the obtained neighboring block as described above, the motion estimation unit is supplied a candidate mode which has high probability of being determined as the mode of the current block. A reference frame data used for encoding the current block and the current block data are inputted to the motion estimation unit. The motion estimation unit performs motion estimation for each candidate mode.

Description

Method and apparatus for motion estimation using mode information of neighboring blocks

The present invention relates to video encoding, and more particularly, to a motion estimation method and apparatus for reducing the time required for motion estimation while maintaining high accuracy.

Recently, H.264 (MPEG4 part 10 / AVC), a video encoding method widely used in mobile television and next-generation DVD, provides a motion compensation block size independent of transform coding size to increase the accuracy of motion estimation. . On the other hand, since the motion vector information must be encoded for each block, the code amount for the motion vector information increases. For this reason, unlike the conventional motion estimation method using fixed-size blocks, the inter mode is divided according to each block size, motion estimation is performed sequentially for all modes, and the coding efficiency is selected by selecting the mode with the lowest cost. Increased.

However, variable block size motion estimation is one of the main factors that increase the complexity of the H.264 / AVC coder. All the motion estimation processes that occupy the largest portion of the video encoding process must be performed for each available mode. Therefore, the amount of computation accordingly increases significantly in proportion to the number of modes. Therefore, it is essential to reduce the amount of computation due to the motion estimation of the variable block size in order to expand the practical application of the H.264 / AVC video standard.

In order to reduce the amount of computation, a method of performing the motion estimation operation only for a specific mode may be possible, but the efficiency is lower than that of the motion estimation operation for all the modes.

SUMMARY OF THE INVENTION The present invention has been made to solve the above problems, and an object of the present invention is a motion estimation method for solving the difficulty of real-time encoder implementation by reducing the computation amount of motion estimation, which takes the largest amount of computation when encoding a video. It is an object to provide a device for.

According to an aspect of the present invention, there is provided a motion estimation method comprising: providing a mode information statistical table containing statistical characteristics of a mode of a current encoding target block according to mode information of a neighboring block that is already encoded; Obtaining mode information of a neighboring block for a current block, and performing motion estimation for a selected candidate mode using the mode information statistics table and mode information of the neighboring block.

The neighboring block may be a left block and an upper block with respect to the current block.

The current block may be a current macroblock to be coded, and the mode information of the neighboring block may include mode information of a right upper 4 * 4 block of the left block and mode information of a lower left 4 * 4 block of the upper block.

The selected candidate mode may include three modes or four modes having a high probability of becoming the mode of the current block among seven modes for motion estimation.

In order to achieve this technical problem, a motion estimation apparatus according to an embodiment of the present invention includes a mode information statistical table containing statistical characteristics of a mode of a current encoding target block according to mode information of a neighboring block that is already encoded, and a current block. And a motion estimator for acquiring mode information of the neighboring block and performing motion estimation on a candidate mode selected using the mode information statistics table and the mode information of the neighboring block.

The neighboring block may be a left block and an upper block with respect to the current block.

The current block is a macroblock currently being encoded, and the mode information of the neighboring block may include mode information of a right upper 4 * 4 block of the left block and mode information of a lower left 4 * 4 block of the upper block.

The selected candidate mode may include three modes or four modes having a high probability of becoming the mode of the current block among seven modes for motion estimation.

As described above, according to the present invention, it is possible to design a low power and low cost video encoder by reducing the computation amount of motion estimation which occupies the most during video encoding.

The above embodiment has been described with reference to H.264, which is one of the video encoding schemes currently widely used. However, this is merely an example, and it should be noted by those skilled in the art that the technical idea of the present invention may be equally applicable to other existing encoding schemes or other standards to be researched and developed in the future.

Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings. For convenience of description, the embodiment will be described based on the H.264 encoder currently widely used. However, this is merely exemplary and the technical spirit of the present invention is not limited thereto.

1 is a block diagram of an H.264 encoder.

Referring to FIG. 1, each macroblock of image data input to an encoder is encoded in an intra mode and an inter mode. Intra mode generates residual macroblocks using neighboring pixel values in the same frame, and in inter mode, residuals are obtained through a motion estimation (ME) and motion compensation (MC) process using a reference frame. ) Create a macroblock. The parameter values and residual macroblocks generated through this process are converted into quantized transform coefficients through integer transform (T) and quantization (Q) and then subjected to entropy coding. At the same time, the generated quantized transform coefficients are subjected to inverse quantization (Q ^-1 ) and integer inverse transform (T ^-1 ), and then added to the values formed in the intra mode or the inter mode. It is used as a reference image reconstructed through a filter process. Among these encoding processes, the present invention relates to motion estimation. 2 is a diagram illustrating a motion estimation mode of H.264.

H.264 supports various inter-mode motion estimation based on variable blocks. Motion estimation using square blocks is also possible, as in mode 1, mode 4, and mode 7, but motion estimation using non-square blocks such as mode 2, mode 3, mode 5, and mode 6 is also possible. The H.264 encoder performs motion estimation for each mode and determines the mode and motion vector with the lowest cost. Since there are various modes as described above, H.264 can achieve a higher compression rate than the previous video encoding scheme. However, since the motion estimation is performed for each mode, the amount of computation required for the motion estimation is very large. The present invention reduces the amount of computation for motion estimation by performing motion estimation for some modes (candidate mode) with high probability instead of performing motion estimation for all modes using statistical characteristics of mode information of neighboring blocks.

3 shows a configuration of a motion estimation apparatus according to an embodiment of the present invention.

The motion estimating apparatus 100 includes a mode information statistics table 110 and a motion estimating unit 120. The mode information statistics table 110 is a table that summarizes the results of experiments on various images under various conditions (resolution, frame rate, Qp value, etc.). The mode information statistics table 110 may be provided separately for each condition or may be provided regardless of the condition. The method of configuring the mode information statistics table 110 will be described later.

The motion estimator 120 acquires mode information of a neighboring block for motion estimation on a block currently being encoded (hereinafter, referred to as a current block). The motion estimation unit 120 receives a candidate mode having a high probability of being determined as the mode of the current block by using the acquired mode information of the neighboring block and the mode information statistics table 110. The motion estimation unit 120 receives data of the current block and data of a reference frame to be used for encoding the current block and performs motion estimation for each candidate mode. The motion estimator 120 outputs residual block data and motion vector information when the lowest cost exists. The motion vector information may be an estimated motion vector value, but may be a difference value between a motion prediction value and the estimated motion vector.

4 is a flowchart illustrating a motion estimation process using the mode information statistics table.

First, the mode information statistics table is provided to the motion estimation apparatus (S10). The mode information statistics table includes information on candidate modes that are likely to be determined as modes of the current block according to mode information of neighboring blocks having a high correlation with the current block.

The motion estimating apparatus obtains mode information of a neighboring block (S20). The mode information of the neighboring block is obtained from the block already encoded. The neighboring block may be at least one block already encoded. For example, the left block and the upper block closest to the current block may be neighboring blocks. In addition to the left block and the upper block, the neighboring block may include a left block (upper block of the left block) of the upper block and a right block of the upper block (right block of the current block).

After acquiring the mode information of the neighboring block, the motion estimation apparatus compares the mode information statistics table with the mode information of the neighboring block (S30), and selects a candidate mode (S40). The candidate mode means some modes having a high probability of being selected as a mode of the current block among seven inter modes. If the candidate mode is selected, the motion estimation apparatus performs motion estimation for each candidate mode (S50). For example, when the number of candidate modes is three, the motion estimation apparatus performs motion estimation on three candidate modes without performing motion estimation on the remaining four modes, and when the number of candidate modes is four. The motion estimation apparatus performs motion estimation for four candidate modes without estimating motion for the remaining three modes.

As mentioned above, in the motion estimation process of various block size modes, there is a correlation between the determined optimal mode of the current block according to the mode of the neighboring block. Therefore, the present invention uses the mode information stored in the mode information statistics table in estimating the current mode according to the mode of the upper block and the left block. This will be described in detail below.

FIG. 5 illustrates a case where a mode information statistics table is configured to determine candidate modes in macroblock units, and FIG. 7 shows a mode information statistics table obtained through experiments on various images.

 When using the existing motion estimation method, motion estimation should be performed for all the macroblocks to be currently encoded from modes 1 to 7, but in this embodiment, prior to performing the motion estimation on the current macroblock, the upper macroblock and Look at the mode information of the left macroblock. Then, the mode information is compared with the mode information statistical table obtained through experiments to select a candidate mode with a high probability.

For example, referring to FIGS. 5 and 7, when the mode of the A block is 1 and the mode of the B block is 1, the mode of the C block is Mode 1, Mode 3, Mode 2, Mode 4, Mode 5, Mode 6 , Mode 7 has the probability of 88.57%, 3.38%, 3.38%, 2.80%, 1.00%, 0.63%, 0.25%.

When the mode of the A block is 2 and the mode of the B block is 3, the mode of the C block is 40.35%, 18.05%, in the order of Mode 1, Mode 4, Mode 2, Mode 3, Mode 6, Mode 5, Mode 7, 15.47%, 14.43%, 5.12%, 4.91%, 1.66%.

For example, in the case where three modes are selected as candidate modes, mode 1, mode 3, and mode 2 become candidate modes when the modes of the A block and the B block are all Mode 1. Considering the case in which four modes are selected as candidate modes, mode 1, mode 3, mode 2, and mode 4 become candidate modes when the modes of the A block and the B block are all Mode 1.

Similarly, when three modes are selected as candidate modes, mode 1, mode 4, and mode 2 become candidate modes when the modes of the A block and the B block are Mode 2 and Mode 3, respectively. Considering the case in which four modes are selected as candidate modes, mode 1, mode 4, mode 2, and mode 3 become candidate modes when modes A and B are modes 2 and 3, respectively.

As described above, the candidate mode may select three or four modes, but it is also possible to select a smaller number of modes or a larger number of modes. In addition, the candidate mode may not be selected as a predetermined number of modes having a high probability, and may be limited to the case where the magnitude of the probability exceeds a specific reference value.

For example, if the candidate mode has a probability of more than 10% and the condition is limited, the candidate mode is only 10% because only mode 1 has a probability of more than 10% when the modes of the A block and the B block are all modes 1. Is selected. On the other hand, when the modes of the A block and the B block are Mode 2 and Mode 3, Mode 1, Mode 4, Mode 2, and Mode 3 become candidate modes.

As described above, the mode information statistics table may be separately configured according to conditions such as image resolution or QP.

5 and 6, the mode information of the A block 20 and the B block 30 is obtained as the mode information of the block 40 and the block 50 which are 4 * 4 blocks, respectively. That is, the mode information of the block A 20, which is the left block of the block C 10, which is the current macroblock, is represented by the mode information of the right upper 4 * 4 block 40 with respect to the block A 20, and the block B 30 is represented. Mode information of) may be represented by the mode information of the lower left 4 * 4 block 50 for the block B (30).

5A illustrates a process of performing motion estimation without using a candidate mode. In this case motion estimation is performed for all modes.

5B shows that motion estimation is performed for three candidate modes except for four modes according to an embodiment of the present invention. In this case, motion estimation is performed for probability 1, 2, and 4 modes.

5C shows that motion estimation is performed for four candidate modes except three modes according to an embodiment of the present invention. In this case, we perform motion estimation for modes 1, 4, and 6 that are probably high.

1 is a block diagram of an H.264 / AVC (hereinafter referred to as H.264) encoder.

2 is a diagram illustrating a motion estimation mode of the H.264 method.

3 is a view showing a motion estimation apparatus according to an embodiment of the present invention.

4 is a flowchart illustrating a motion estimation process according to an embodiment of the present invention.

5 is a diagram illustrating a case of configuring a mode information statistics table according to an embodiment of the present invention.

FIG. 6 is a diagram illustrating a case in which mode information about a neighboring block is represented by mode information of a 4 * 4 block.

7 is a diagram illustrating an example of a mode information table obtained by using various conditions and various images.

8A to 8C are diagrams illustrating an example of a motion estimation process using a candidate mode in which some inter modes are removed.

Claims (9)

Providing a mode information statistics table containing statistical characteristics of a mode of a current encoding target block according to mode information of an already encoded neighboring block; Obtaining mode information of a neighboring block for the current block; And And performing motion estimation for the selected candidate mode using the mode information statistics table and the mode information of the neighboring block. The method of claim 1, The neighboring block is a motion estimation method using mode information of the neighboring block, characterized in that the left block (upper block) and the left block (upper block) for the current block. The method of claim 2, The current block is a macroblock currently being encoded, and the mode information of the neighboring block includes mode information of a right upper 4 * 4 block of the left block and mode information of a lower left 4 * 4 block of the upper block. Motion estimation method using mode information of neighboring blocks. The method of claim 1, The selected candidate mode includes three modes or four modes having a high probability of becoming the mode of the current block among the seven modes for motion estimation. A mode information statistics table containing statistical characteristics of a mode of a current encoding target block according to mode information of an already encoded neighboring block; And A motion using mode information of a neighboring block including a motion estimation unit for acquiring mode information of a neighboring block for a current block and performing motion estimation for a selected candidate mode using the mode information statistics table and the mode information of the neighboring block. Estimation device. The method of claim 5, The neighboring block is a motion estimation device using mode information of the neighboring block, characterized in that the left block and the upper block for the current block. The method of claim 6, The current block is a macroblock currently being encoded, and the mode information of the neighboring block includes mode information of a right upper 4 * 4 block of the left block and mode information of a lower left 4 * 4 block of the upper block. Motion estimation device using mode information of neighboring blocks. The method of claim 5, The selected candidate mode includes four modes having a high probability of becoming the mode of the current block among the seven modes for motion estimation. A medium having a program readable by a computing device for executing the method according to any one of claims 1 to 4.

Images