JPH08237668A - Motion vector detection method - Google Patents

Abstract

PURPOSE: To efficiently detect a motion vector with extremely high detection accuracy by performing block matching by a minimum detection accuracy unit in the range of peripheral ±a picture elements based on a first motion vector. CONSTITUTION: Candidate blocks 62 are set within a detection range 60 on reference pictures with the interval of α picture elements respectively in horizontal and vertical directions and the block matching of an encoding unit block 64 and all the set candidate blocks 62 is performed. From the result of performing the matching, the position relation of an encoding object block and the candidate block for minimizing an error is defined as the first motion vector 66. With the candidate block of a position displaced for the first motion vector 66 as a center, ±a picture elements both in horizontal and vertical directions are turned to the detection range and full search is performed by the minimum unit of motion vector detection accuracy. However, for the candidate block for which the matching is performed at the time of detecting the first motion vector 66, the need of performing the matching again is eliminated since judgement is already ended.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a moving picture coding system in a moving picture communication apparatus which is used by connecting to a relatively low speed line such as a telephone line. The present invention relates to a motion vector detection method performed.

[0002]

2. Description of the Related Art Coding of a moving image that is generally used will be described with reference to the drawings. The frame described in the text is one processing unit of an image and includes a so-called field image. In FIG. 3, the image input signal 200 is divided into blocks having a block size of the number of pixels determined by the block divider 201. In the changeover switch 202, in the frame (INTR
A) Encoding or switching between frames (INTER) encoding is performed based on the determination of the encoding control unit 205, but here, only INTER encoding using motion compensation prediction, which is a feature of moving image encoding, will be described. However, the INTRA encoding that does not use the motion compensation prediction will be omitted.
The subtracter 211 obtains the difference between the blocked input signal and the motion-compensated predicted image of the decoded image of the previous frame stored in the frame memory 208, and sets it as the prediction error image. The prediction error image is output as a quantized output 217 of the transform coefficient by the orthogonal transformer 203 and the quantizer 204. The predicted image is generated by the local decoder 219 including the inverse quantizer 206, the inverse orthogonal transformer 207, the adder 210, the frame memory 208, and the motion compensation unit 209. The frame memory 208 stores the decoded image of the previous frame, and the stored image and the current image input signal blocked are input to the motion vector detector (212, 213) to refer to the decoded image of the previous frame. A motion vector is detected as an image. The motion compensation unit 209 performs motion compensation based on the detected motion vector to generate a predicted image of the current frame.

A conventional motion vector detecting method using block matching will be described below. Figure 4 here
The block used for matching as shown in
A rectangular block 32 of 6 pixels and 16 pixels in the vertical direction is used, and the motion vector detection range 30 is ± 1 in both the horizontal and vertical directions.
The area 30 has 5 pixels. Typical motion vector detection methods include full search, step search, and two-step search, which will be described in order with reference to the drawings.

1. Full Search In this method, the candidate blocks set on the decoded image (reference image) of the previous frame are sequentially displaced pixel by pixel in the region of the detection range 30, and block matching is performed with the target block. The matching with all candidate blocks is completed, and the positional relationship with the candidate block with the smallest error is obtained. For the error evaluation for obtaining this positional relationship, for example, the sum of absolute values of differences between corresponding pixels of both blocks is often used. The displacement amount to the candidate vector position where this error is the minimum is set as the optimum motion vector. The number of times of block matching per target block is 961, which is the image format QCIF (Q
uarter Common Intermediate Format 176 × 14
4 pixels), the number of matching times per frame is 9
5,139 times.

[0005] 2. Step search This is one of the hierarchical motion vector detection methods, and as shown in FIG. 5, it is set at a position corresponding to the encoding target block 40 and the motion vector shifted ± 8 pixels in both horizontal and vertical directions on the reference image. Matching is performed with the eight candidate blocks 42 described above, and the positional relationship with the candidate block 43 having the smallest error is obtained. ± 4 from the next calculated position 43
Matching is performed with the eight candidate blocks 45 shifted by pixels, and further matching is performed with the eight candidate blocks 47 shifted by ± 2 pixels. Finally, matching is performed with eight candidate blocks around the reference image shifted by ± 1 pixel, and the obtained result is used as a motion vector to be obtained. The number of block matching per target block is 33 times,
The number of matching times per frame is 3,267. This is about 1/3 compared to the processing by full search
It will be 0. 3.2 Two-step search An explanatory diagram of this method is shown in FIG. Encoding target block 5
The setting of the candidate block 52 for matching 0 with the reference image is arranged every 3 pixels over the entire detection range 30, and the positional relationship with the candidate block 54 having the smallest error is obtained. Furthermore, matching is performed with eight candidate blocks 54 (black circles in FIG. 6) corresponding to the motion vector displaced by ± 1 pixel from the surroundings, and the positional relationship with the candidate block having the smallest error is determined as the final motion vector. To do.
The number of block matchings per target block by this method is 129, and the number of block matchings per frame is 12,771. This is about 1/7 of the processing amount of full search.

[0006]

Each of the conventional motion vector detecting methods described above has the following problems. Motion vector detection by full search has high detection accuracy and can reduce prediction error. As a result, the amount of generated code can be suppressed, so that the image quality of the decoded image on the receiving side can be improved. However, on the other hand, since the number of times of block matching is large and the processing amount becomes huge, in a moving image communication device at a low transmission rate such as a telephone line,
An increase in the load of motion vector detection processing causes a coding delay.

On the other hand, in the motion vector detection by the step search, the number of times of matching is reduced and the processing amount is reduced, but since only some of the candidate blocks are matched, the false detection of the motion vector is likely to occur. Especially when the movement between frames is large, the probability of false detection is high,
Prediction error increases due to a decrease in motion compensation accuracy, resulting in an increase in the amount of generated codes, which causes image quality deterioration in a device with a limited communication rate. Further, in the two-step search, the processing amount and the detection accuracy are intermediate between the above two methods. However, if there is an edge or an independent area in the target block, it is not always 1
Since the search in the second stage does not necessarily detect the optimum motion vector itself or the motion vector adjacent to the optimum motion vector, only the eight candidate blocks in the vicinity are detected in the search in the second stage. False detections will occur if not used. As a result, the prediction error increases as in the case of step search, which causes deterioration of the image quality of the decoded image.

[0008]

SUMMARY OF THE INVENTION The present invention is intended to solve these problems, and a preceding frame is provided for each coding unit block in which a prescribed number of adjacent pixels of a digitized image input signal are collected. In a motion vector detection method for detecting a positional relationship with a current frame from difference information between a decoded image and a corresponding block as a reference image, a motion vector detection range of the reference image is equalized with a predetermined pixel interval α. A first motion vector detector for performing block matching with a plurality of candidate blocks arranged in a plurality of candidate blocks, and a range of surrounding ± α pixels based on the first motion vector detected by the first motion vector detector Of all candidate blocks except the candidate block that has been block-matched by the first motion vector detector in the The second performs a block matching Te
And a second motion vector detector for detecting the second motion vector as a motion vector, and an optimum motion vector is detected based on the second motion vector.

[0009]

According to the motion vector detecting method configured as described above, it is possible to reduce the processing amount of motion vector detection when performing inter-frame coding of a moving image, and to detect various motions. The detection accuracy is high, and efficient motion vector detection is possible.

[0010]

DESCRIPTION OF THE PREFERRED EMBODIMENTS The present invention will be described below with reference to the drawings. FIG. 1 is a flow chart showing a first embodiment of the present invention, and FIG. 2 is an explanatory diagram showing motion vector detection according to the first embodiment of the present invention. The motion vector detection operation will be described with reference to FIG. An image signal is input (step 1, hereinafter referred to as S1), divided into n coding target blocks each having a prescribed number of pixels (S2), and motion vector detection (S4 to S
Perform 7). Block matching is used to detect the motion vector. In block matching, the decoded image of the previous frame is used as a reference image to find the positional relationship with a candidate block that minimizes the error from the target block. For the evaluation of this error, the sum of absolute values of the differences between the pixels corresponding to both blocks is often used, but other evaluation methods such as evaluation by the sum of squares of the differences may be used. In this embodiment, the image size of one frame is QCIF (176 × 1).
44 pixels), and as shown in FIG. 4, one encoding unit block is a rectangular block 32 consisting of 16 pixels in the horizontal direction and 16 pixels in the vertical direction, the detection range of the motion vector in the horizontal direction,
The area 30 of ± 15 pixels in the vertical direction will be described. First, candidate blocks 62 are set in the detection range 60 on the reference image in the horizontal and vertical directions at α pixel intervals (S4). In the example shown in FIG. 2, α is set to 3, and the candidate block 62 is set to all 121 blocks corresponding to the following motion vectors.

(-15, -15), (-12, -15), ... (+15, -15) (-15, -12), (-12, -12), ... ( +15, -12) ・ ・ ・ ・ ・ ・ ・ ・ ・ ・ ・ ・ ・ ・ ・ ・ ・ ・ ・ ・ ・ ・ ・ (-15, +15), (-12, +15), ・ ・ ・-(+ 15, +15) However, (horizontal displacement amount, vertical displacement amount) is displayed. Block matching is performed with the coding unit block 64 and all of the set candidate blocks 62 (S5). Based on the result of the matching, the positional relationship between the candidate block having the smallest error and the encoding target block is set as the first motion vector 66 (S6). With the candidate block at the position displaced by the first motion vector 66 as the center, ± α pixels are set as the detection range in both the horizontal and vertical directions, and the full search is performed in the minimum unit of motion vector detection accuracy (S7). However, since the determination has already been completed for the candidate block that has been matched when the first motion vector 66 is detected,
There is no need to match again. In the example of FIG. 2, α is 3
Therefore, the area of 7 × 7 pixels is set as the detection range 68.
For example, if the first motion vector 66 is (X1, Y1), the range corresponding to the following motion vector is set as the detection range 68.

(X1-3, Y1-3), ... (X1 + 3, Y1-3) (X1-3, Y1-2), ... (X1 + 3, Y1-2) ··············· (X1-3, Y1 + 3), ··· (X1 + 3, Y1 + 3) Here, the minimum unit of motion vector detection is 1 In terms of pixels, the setting of candidate blocks to be block-matched with the block to be encoded is carried out from all 49 possible candidate blocks in the detection range 68 to 9 candidate blocks that have been matched at the time of detection of the first motion vector 66. All 40 except
Make it a candidate block. The black dots 70 in FIG. 2 indicate the positions of the motion vectors corresponding to the 40 candidate blocks. From the result of matching with the 40 candidate blocks, the second positional relationship with the candidate block having the smallest error is determined.
(S8). The determination of INTRA / INTER is performed using the second motion vector (S
9). This is a coding mode determination as to whether the target block 64 is to be coded by intra-frame coding (INTRA) or inter-frame coding (INTER). When it is determined that the coding is performed in the INTER mode, the value of the second motion vector is adopted as the optimum motion vector (S10) and used for motion compensation prediction. The above series of processing is performed for all the blocks in one screen, and when all the blocks are completed, the process moves to the next screen.
Next, the number of block matchings per block to be encoded in the present embodiment is calculated to be 161 times, and it is 15,771 times per frame. This is a calculation amount of about 1/6 as compared with the matching by the full search. In addition, the amount of processing is about 5 times that of the step search. However, in the setting of the candidate blocks of this embodiment, no detection omission occurs, so that the detection accuracy of the obtained motion vector is almost the same as the detection by full search, and the occurrence of prediction error can be significantly suppressed. As a result, even in the coding of moving image communication when the communication speed such as a telephone line is low and limited, the generated code amount is suppressed, so that a decoded image with good image quality can be obtained on the receiving side. In this embodiment, the minimum pixel unit for motion vector detection has been described as one pixel, but in recent moving image coding, the image quality is improved in half pixel units. According to the present invention, it is possible to apply to such motion vector detection in units of half pixels.

[0013]

As described above, according to the method of the present invention, the amount of processing required to detect a motion vector can be reduced, and the detection accuracy can be improved even when detecting various motions such as motions of edges and independent regions. Extremely high and efficient motion vector detection is possible. As a result, since the occurrence of prediction error is suppressed to a low level, the amount of generated code is suppressed even in the case of image communication coding where the communication speed is low and limited, such as a telephone line. A good decoded image can be obtained.

[Brief description of drawings]

FIG. 1 is a flowchart showing an embodiment of the present invention.

FIG. 2 is an explanatory view showing an embodiment of the present invention.

FIG. 3 is a block diagram of a moving image encoder.

FIG. 4 is an explanatory diagram showing a motion vector detection range.

FIG. 5 is an explanatory diagram showing a step search.

FIG. 6 is an explanatory diagram showing a two-step search.

[Explanation of symbols]

 200 image input signal 201 block divider 202 INTRA / INTER switch 203 orthogonal transformer 204 quantizer 205 coding controller 206 inverse quantizer 207 inverse orthogonal transformer 208 frame memory 209 motion compensator 210 adder 211 subtractor 212 first motion vector detector 213 second motion vector detector 219 local decoder 30 motion vector detection range 32, 40, 50, 64 block to be encoded

─────────────────────────────────────────────────── ───

[Procedure amendment]

[Submission date] November 6, 1995

[Procedure Amendment 1]

[Document name to be corrected] Drawing

[Name of item to be corrected] Figure 1

[Correction method] Change

[Correction content]

FIG.

Claims (1)

[Claims] 1. A positional relationship with a current frame based on difference information from a corresponding block using a preceding frame decoded image as a reference image for each coding unit block in which a prescribed number of pixels close to each other of a digitized image input signal are collected. In the motion vector detecting method for detecting the motion vector, a first motion vector for performing block matching with a plurality of candidate blocks evenly arranged at a predetermined pixel interval α within the motion vector detecting range of the reference image. A detector and a candidate block subjected to block matching by the first motion vector detector within a range of surrounding ± α pixels based on the first motion vector detected by the first motion vector detector. To detect the second motion vector by performing block matching with all candidate blocks except Motion vector detecting method is composed of a second motion vector detector, optimal motion vector based on the second motion vector is characterized in that it is detected.