JPS61201584A - Dynamic vector detector - Google Patents

Abstract

PURPOSE:To prevent effectively the reduction in coding efficiency due to ineffective dynamic vector and deterioration in resolution by forecasting the dynamic vector of present frame from the entire movement of an animation picture and correcting a detected dynamic vector exceeding largely the forecast value. CONSTITUTION:A dynamic vector forecast section detects dynamic vectors of three consecutive frames before the present frame via one-frame delay circuits 3a-3c connected in cascade by 3 stages and the dynamic vector of the present frame is forecast as the weight means of the dynamic vector. A differential device 7 obtains a difference DELTAu1 between the dynamic vector u1 detected at the present frame and the dynamic vector u1 forecast at a forecast section as an absolute difference. A discriminator 8 compares the dynamic vector difference DELTAu1 with a preset limit L and when DELTAu1>L, it is discriminated as an ineffective vector because the detected dynamic vector u1 of the present frame is unnatural as the substantial movement of the animation picture.

Description

DETAILED DESCRIPTION OF THE INVENTION [Technical Field of the Invention] The present invention relates to a motion vector detection device that can correctly detect motion vectors in moving image processing.

[Technical background of the invention and its problems]

When transmitting TV images or movie film images, the bandwidth is reduced by one in order to reduce the transmission cost. As one of the band compression techniques, an attempt has been made to obtain the parallel shift t' between frames of a moving image and obtain the motion vector, thereby performing motion compensation frame and inter-coding, and increasing the coding compression rate. ing.

In addition, as shown in Figure 3, the motion part and total correlation calculation between temporally consecutive frames is performed, the amount of parallel movement is detected as a total motion vector, and this is used to perform interframe interpolation processing to obtain a high It is performed to generate a moving image with a resolution of τ.

However, conventionally, a motion vector is obtained by detecting the amount of parallel movement between each frame and the previous frame. That is, motion vectors are determined independently for each frame. For this reason, if the motion vectors detected in each frame exhibit changes as shown in FIG. 4, for example, if all of them are used for motion compensated interframe coding, efficient coding will not be possible. Furthermore, if all of this is used as is for interframe interpolation, a significant deterioration in the image quality of the reproduced image will result. In other words, considering that there is a large correlation between temporally consecutive frames of a moving image, and that the change is slow compared to the frame period, for example, a phenomenon such as that caused by shake of a TV camera as shown in Figure 4 can be considered. It can be said that the change in motion vector does not represent the original motion of the moving image.

Therefore, if such invalid motion vectors are used as they are in moving image processing, this causes a significant deterioration in image quality.

[Purpose of the invention]

The present invention was made in consideration of these circumstances, and its purpose is to easily and effectively detect motion vectors effective for video image processing, and to enable image reproduction without deterioration of image quality. An object of the present invention is to provide a motion vector detection device that does the following.

[Summary of the invention]

When detecting all the motion vectors of the current frame from the correlation of 27 consecutive norms, the present invention predicts the motion vectors of the current frame from the motion vectors detected in frames before the current frame, and combines this predicted motion vector with the current frame. The motion vector detected in the current frame is output as valid only when the difference with the motion vector detected in is smaller than a predetermined limit value, and the difference between the vectors is larger than the limit value. Sometimes, the motion vector detected in the current frame is modified and output.

That is, when the difference between the predicted motion vector and the detected motion vector is larger than a limit value, it is determined that the detected motion vector is invalid, and for example, the detected motion vector is output as zero, or the image is When the motion vector is detected by dividing the image into regions, the motion vector is determined by referring only to the vectors in the valid region, and the motion vector is corrected.

〔Effect of the invention〕

Thus, according to the present invention, the motion vector of the current frame is predicted from the overall motion of the moving image, and detected motion vectors that deviate significantly from the predicted value are corrected, so that only motion vectors that are effective for moving image processing are efficiently processed. It becomes possible to detect it well.

As a result, it becomes possible to efficiently perform moving image processing such as motion compensated interframe coding and interframe interpolation using the frame period, and to improve image quality. In other words, great practical effects can be achieved, such as effectively preventing a reduction in encoding efficiency and deterioration in resolution due to invalid motion vectors.

[Embodiments of the invention]

Embodiments of the present invention will be described below with reference to the drawings.

FIG. 1 is a schematic configuration diagram of an embodiment device. A motion vector detection circuit (detection device main body) 1 calculates the motion vector υ in the current frame from the correlation between two consecutive frames of a moving image.
It detects everything. Typical methods for detecting motion vectors include a method in which all motion vectors are detected for the entire frame screen as one region, and a method in which motion vector candidates are obtained for each region in which the entire screen is divided into multiple regions, and motion vector candidates are calculated from these motion vector candidates. Although there is a method of determining the motion vector of the entire screen, the explanation here assumes that the former method is used.

The motion vector detected for each frame by the motion vector detection circuit 1 is input to the selector circuit 2, as well as to a motion vector prediction section and a motion vector determination section. The motion vector prediction unit here includes l-frame delay circuits 3a p 3b a 3 connected in cascade in three stages.
All motion vectors of three consecutive frames before the current frame are detected through e', and all motion vectors of the current frame are predicted as a weighted average of these motion vectors.

That is, when the motion vector detected in the current frame is vn, the delay circuits Ja, Jb, 3c calculate the motion vector n-1n-2'τ of three consecutive frames before the current frame.
Find n-3 and obtain the predetermined coefficients α, β via the coefficient circuit vector τ, V 4h, 4b, 4cf.

After each γ is multiplied, an adder 5 calculates the added value (α−vn−1+β·ga, +γ·vn−3). Then, this added value is quantized via the quantizer 6 to obtain the predicted motion vector τ1 of the current frame, for example, v1''(α'''r)n-1+β''n-2+γ''n-5
).

The motion vector determination unit includes a subtractor 7 and a determiner (comparator) 8, and the subtractor 7 calculates the motion vector detected in the current step and the motion vector predicted by the prediction unit. The difference Δτ1 from the motion vector 4 is determined as the absolute value difference Δτ1=1τ1−1. The determiner 8 compares the motion vector difference Δτ1 with a preset limit value, and if Δτ1<L, the detected motion vector v1 of the current frame is determined.
If it is determined that Δτ,〉L is valid, then the detected motion vector τ of the current frame has changed significantly beyond its predicted value, and the motion is considered unnatural as the original motion of the moving image. It has been judged invalid.

The selector circuit 2II'i receives the output of the above-mentioned determiner 8, and if the determination result is 1 valid 1, the motion vector τ, t- of the current frame determined by the motion vector detection circuit 1 is determined. It is output as is. If the determination result is "invalid", "0" is selected and output in place of the detected motion vector τ of the current frame.

This motion vector '0' has the meaning of marking the motion vector υ1 detected in the current frame as invalid and correcting it as if there was no motion vector in that frame.

In addition, in the case of the method of determining the overall motion vector from the motion vector candidates obtained from each of the multi-divided areas as described above, for example, the direction opposite to the overall motion of the moving image is The correction process may be performed to invalidate the vector component indicating the motion of the motion vector and determine the overall motion vector by referring only to the remaining valid motion vector candidates.

According to the present apparatus configured as described above, for example, when the detected motion vector of each frame exhibits random changes due to shake of a TV camera, etc., undesirable changes may occur from the perspective of the overall motion flow of the moving image. Since all motion vectors are invalidated and corrected, it is possible to obtain only motion vectors that are effective for motion compensation interframe coding processing and interframe interpolation.

Therefore, it is possible to prevent deterioration in encoding efficiency, deterioration in image quality, etc., and to perform good moving image processing.

Note that the present invention is not limited to the embodiments described above. For example, in the embodiment, the motion vector of each frame detected by the motion vector detection circuit 1 is used as is to predict the motion vector of the current frame. However, as shown in FIG. The configuration may be such that all motion vectors of the current frame are predicted from the motion vector). Furthermore, although here, the motion vectors of three frames were used to obtain the predicted motion vector, motion vectors of four or more consecutive frames may be used. It is also possible to employ prediction methods other than weighted average. Furthermore, the method of correcting the motion vector can be modified as appropriate. In short, the present invention can be modified in various ways without departing from the gist thereof.

[Brief explanation of drawings]

FIG. 1 is a schematic block diagram of a device according to an embodiment of the present invention, FIG. 2 is a schematic diagram of a device according to another embodiment of the present invention, and FIGS. 3 and 4 illustrate problems in conventional motion vector detection. It is a diagram for explaining. 1...Motion vector detection circuit, 2...Selector circuit1. ? a, 3b, Jc-17 frame delay circuit, 4a, 4
b, 4c...Coefficient unit, 5...Adder, 6...Quantizer, 7...Differentiator, 8...Determiner.

Claims (3)

[Claims] (1) A detection device that detects the motion vector of the current frame from the correlation between two consecutive frames; a means for predicting the motion vector of the current frame from the motion vector detected in the frame before the current frame; means for determining a difference between a motion vector and a motion vector detected in the current frame, and validating the detected motion vector when the difference between the motion vectors is smaller than a predetermined limit value and detecting the detected motion vector when the difference is larger than a predetermined limit value; 1. A motion vector detection device comprising: means for correcting a motion vector. (2) The motion vector detection device according to claim 1, wherein the correction of the detected motion vector is to uniquely set the motion vector to zero. (3) A patent claim in which the detected motion vector is corrected by determining the overall motion vector by referring only to valid vector candidates among the motion vector candidates obtained for each multi-divided region. The motion vector detection device according to item 1.