JPH0543681U - Image coding device - Google Patents

Abstract

(57) [Summary] An image coding apparatus that increases search speed by reducing search points when performing motion compensation prediction. [Object] To improve the encoding processing speed of an image encoding device. [Structure] When performing motion-compensated prediction, the previous motion vector is memorized for each block, the direction of the motion vector in the next frame is estimated from that value, and the place with low probability of motion vector movement Do not calculate and reduce the number of search points. [Effect] By eliminating the calculation at a search point having a low probability of movement, the time required for encoding is reduced, and high-speed encoding processing can be performed.

Description

[Detailed description of the device]

[0001]

[Industrial applications]

 The present invention relates to an encoding unit of an image encoding device.

[0002]

[Prior Art]

 FIG. 6 shows, for example, “NEC Technical Report” Vol. 41 No. FIG. 9 is a diagram showing motion vector detection in motion compensation prediction of a conventional image coding device shown in 9/1988 “TV signal coding technology”. 1 is the origin of the block, 2 is the search point with the highest correlation in the first stage, 3 is the motion vector in the first stage, 4 is the search point with the highest correlation in the second stage, and 5 is the second The motion vector at the stage, 6 is the search point with the highest correlation at the third stage, 7 is the motion vector at the third stage, and 8 is the final motion vector.

Next, the operation will be described. The correlation between the block at the origin 1 of the motion-compensated prediction of the current frame and the block at the search points 9 (white points) in the first stage of the previous frame is calculated, and the search point with the highest correlation is set to 2, The vector from the origin 1 to the search point 2 is 3. Next, centering on the search point 2, find the correlation with the block at the origin of the current frame for each of the 8 search points (black points) that are more accurate than in the first step, and set the search point with the highest correlation to 4. I will. The vector from search point 2 to search point 4 is 5. Next, for each of the 8 search points (X points) that are more accurate than in the second stage, centered around the search point 4, find the correlation with the block at the origin of the current frame, and find the search point with the highest correlation. will do. The vector going from search point 4 to search point 6 is set to 7. The sum of vector 3, vector 5 and vector 7 is the motion vector of this block, and this is 8. In this way, the motion vector is obtained.

[0004]

[Problems to be solved by the device]

 Since the conventional image coding apparatus is configured as described above, it is necessary to perform calculation for search points in all directions in the first search, and it is obvious that the search points will not move. On the other hand, there was a problem that calculation had to be performed.

The present invention has been made to solve the above-mentioned problems, and in the motion compensation prediction, the calculation is not performed for the search points that apparently do not move in the next frame, and the calculation time is reduced. The purpose is to shorten.

[0006]

[Means for Solving the Problems]

 The image coding apparatus according to the present invention stores a previous motion vector for each block for which motion compensation prediction is performed, and judges from that memory, a search point with a low probability that the motion vector in the next frame will move. We decided to stop calculating the correlation at.

[0007]

[Action]

 The image coding apparatus according to the present invention stops the calculation of the correlation at the search point where the probability of movement is low, thereby reducing the amount of calculation required for motion compensation prediction.

[0008]

【Example】

 Example 1. An embodiment of this invention will be described. In FIG. 1, 9 is a frame memory that stores the data of the previous frame, 10 is a portion that performs motion compensation from the data of the previous frame and the data of the current frame, 11 is a subtractor that performs the difference between the input data and the prediction data, 12 Is a quantizer for compressing and quantizing difference data, 13 is an inverse quantizer for dequantizing quantized data, 14 is an adder for adding prediction data and difference data, and 15 is obtained by motion compensation. It is a memory for storing the motion vector.

The operation will be described below. The input data is subtracted from the prediction data by the subtractor 11, and the difference data is quantized by the quantizer 12 and transmitted. On the other hand, the quantized data is dequantized by the dequantizer 13 and added to the prediction data by the adder 14 and stored in the frame memory 9. As to how the prediction data is formed, the motion compensation unit 10 checks the correlation between the input data and the previous frame data to obtain the motion vector. The motion vector is stored in the memory 15, and the portion obtained from the motion vector from the frame memory is output as prediction data.

In FIG. 2, 1 is the origin of a unit block in motion compensation prediction, 2 is the search point with the highest correlation in the previous motion compensation, and 3 is the previous motion vector (search point 2 from the origin of the current frame to the previous frame). Heading to). The points are all search points, and the black points are the search points for which the correlation is calculated this time.

First, it is assumed that there is an existing unit block in motion compensation prediction, and the previous motion vector (from the origin of the current frame to the search point 2 of the previous frame) is 3. That is, it is assumed that the image is moving from the search point 2 to the search point 1. This motion vector (horizontal component, vertical component) is stored in memory.

Here, in order to perform motion compensation prediction for the next frame, the sum of absolute differences between pixel data between the origin and each search point or the sum of squared differences is calculated to find the search point with the highest correlation. Explore.

However, since the image moved from the search point 2 to the search point 1 last time, the probability that the image moves in the same direction in the next frame is very high. Therefore, only the black search points that are almost close to the direction from the search point 2 to the search point 1 will be searched and calculated.

That is, only the search points in the area where the sign of the previous motion vector is the same (+ if +, − if −) are searched.

FIG. 3 is a flowchart of this operation. First, in step 16, the motion vector is determined, and in step 17, the motion vector is stored in the memory. In step 18, the range to be searched from the motion vector is determined, and in step 19, the correlation at the search point within the range is calculated.

Example 2. FIG. 4 shows another embodiment, in which 1, 2, 3 are the same as those in FIG. 2, and black points, that is, quadrants in which the search point for calculation is in the opposite direction of the motion vector are excluded. It is within the range. The operation is the same as in the previous embodiment, and the search range is wide, so it is considered to be effective for images with a small number of frames.

Example 3. FIG. 5 is a flowchart showing another embodiment. In step 20, the motion vector is determined, and in step 21, the motion vector is stored in the memory. In step 22, the magnitude of the motion vector is calculated. If it is larger than the threshold value, a search is performed in the same range as in FIG. 2 in step 23, and if it is smaller than the threshold value, a full search is performed in step 24. In step 25, the correlation is calculated. This is considered to be more effective than the first embodiment because the greater the movement, the greater the probability that the next movement direction will be the same.

[0018]

[Effect of the device]

 As described above, according to the present invention, among the search points in the motion-compensated prediction, the points with a low probability of movement are deleted, so the amount of pixel calculation for the block is reduced, the time required for motion compensation is reduced, and the coding is performed. It has the effect of reducing the processing time.

[Brief description of drawings]

FIG. 1 is a diagram showing a configuration of an encoding unit according to an embodiment of the present invention.

FIG. 2 is a diagram showing motion vector search points according to an embodiment of the present invention.

FIG. 3 is a flow chart showing the operation of this invention.

FIG. 4 is a diagram showing motion vector search points according to another embodiment of the present invention.

FIG. 5 is a flow chart showing the operation of another embodiment of the present invention.

FIG. 6 is a diagram showing search points of a conventional device of this type.

[Explanation of symbols]

 9 frame memory 15 memory for storing motion vectors

Claims (1)

[Scope of utility model registration request] 1. An image coding apparatus that divides input image data into blocks and performs motion compensation prediction, includes a frame memory that stores data of a previous frame, and a memory that stores a previous motion vector for each block. Image coding device.