JPH01177785A - Dynamic vector detecting method for image signal - Google Patents

Abstract

PURPOSE:To detect dynamic vector of high reliability even when an accurate dynamic vector is hard to be detected by the influence of noises or the nature of an image, etc., by providing a reliability deciding means to decide the reliability of a detected dynamic vector. CONSTITUTION:In the process of detecting the dynamic vector from a supplied image signal, a step in which the reliability of a detected dynamic vector is decided is provided; an image is divided into blocks in fixed size, and dynamic vector is detected for each of the blocks. An image signal is inputted through a frame memory 9, a variable delay circuit 10, and a differentiator 12 to an evaluation value calculating circuit 13, where a value for each searching vector in each block is calculated. And through a memory 14, a vector detection circuit 15, and the reliability detection circuit 16, the reliability is decide, and the result is supplied to a dynamic vector determining circuit 17. In this case, such a block in which no highly reliable dynamic vector is found is expanded in size, so that a highly reliable dynamic vector can be detected even when an accurate dynamic vector can hardly be detected.

Description

DETAILED DESCRIPTION OF THE INVENTION (Field of Industrial Application) The present invention relates to a motion vector detection technique for image signals.

(Prior Art) First, motion vectors will be explained using FIG. 6. The figure shows a case where a triangle is moving on the image.

A motion vector is the direction and amount of movement of a moving object on an image between different frames in the time direction of an input image signal, for example, the current frame and the previous frame.

Conventionally, a supplied image is divided into blocks of a predetermined size, and for each block, an evaluation value,
For example, the sum of absolute values of inter-frame difference values within a block or 2
A commonly used method is to calculate the sum of the products and use the best evaluation result as the motion vector.

(Problem to be solved by the invention) However, when detecting the motion vector of the input image,
In the conventional method, an evaluation value is obtained using a predetermined method for each vector to be searched, and the best evaluation result is used as a motion vector. This method has the disadvantage that vectors that correctly indicate the movement of the subject on the image cannot necessarily be detected due to the influence of noise. An object of the present invention is to detect a highly reliable motion vector even when it is difficult to detect an accurate motion vector due to the influence of noise or the properties of an image.

(Means to solve the problem).

According to the present invention, when detecting a motion vector from a supplied image signal, there is a step of determining the reliability of the detected motion vector, dividing the image into blocks of a predetermined size, and dividing the image into blocks of a predetermined size. Perform motion vector detection for
For blocks whose detected vectors are not determined to be highly reliable in the reliability determination step, the block size is expanded according to a predetermined method, and the expanded block is detected as a motion vector again. If a highly reliable vector is detected, that vector is output as a detected motion vector in a block of the initially set size, and if a highly reliable vector is not detected. In this case, a motion vector detection method is obtained in which a series of operations of enlarging the block size, detecting a motion vector, and determining the reliability of the detected vector are further repeated.

(Operation) When dividing the supplied image signal into blocks of a predetermined size and detecting motion vectors for each block, for blocks for which a highly reliable motion vector was not found, the size of that block is determined. By increasing the size, it becomes possible to obtain a highly reliable motion vector.

The principle of the present invention will be explained below with reference to FIG.

The image input unit 1 stores two frames separated by a predetermined interval from the supplied image signal.

The block size initial setting unit 2 sets the block size for calculating the evaluation value of a vector to a predetermined size.

The evaluation value calculation unit 3 calculates evaluation values for each of a plurality of search vectors in a predetermined range in a set block size.

The motion vector detection section 4 detects the vector with the best evaluation value calculated by the evaluation value calculation section 3 as a motion vector. If necessary, a method may be adopted in which several vectors are extracted in order from the vector with the best evaluation value.

The reliability determination unit 5 determines the reliability of the detected motion vector, and if it is determined that the reliability is high, the process branches to the detected motion vector output unit 8, and if it is determined that the reliability is low. If so, the process proceeds to the vector detection stage number determination unit 6.

In the vector detection stage number determining unit 6, if the preset number of vector detection stages has been reached, the process branches to the detected motion vector output unit 8; otherwise, the process proceeds to the block size enlarging unit 7.

The detected motion vector output unit 8 outputs the detected motion vector as a motion vector for the initially set block. In the case of a branch from the vector detection stage number determining unit 6, a method may be used in which information indicating that a highly reliable vector has not been detected or a zero vector is output.

The block size enlarging section 7 enlarges and resets the block size for calculating the evaluation value of the vector using a predetermined method, and returns to the evaluation value calculating section 3.

The effect on vector detection when increasing the block size when calculating the evaluation value of a vector will be described below with reference to FIGS. 2 and 3.

FIGS. 2 and 3 show the evaluation values of each search vector obtained for one of the blocks by dividing the image into blocks of a certain size. Furthermore, the block size in FIG. 3 is larger. The two horizontal axes indicate the horizontal and vertical directions of the search vector in the image, and the vertical axis indicates the evaluation value when using that vector, here the sum of squares of the interframe prediction error within the block. There is. That is, according to the conventional method, the vector with the smallest evaluation value is detected as the motion vector.

Normally, when calculating the evaluation value to detect a motion vector, the result is a large wavy state as shown in Figure 2 due to the influence of noise contained in the image signal. There are many vectors for which . Therefore, the vector with the smallest evaluation value is not necessarily the vector that correctly indicates the movement of the subject on the image. On the other hand, if you increase the size of this block, the noise has no correlation between adjacent pixels, so its energy is not proportional to the proportion of the block size, and the influence of noise becomes relatively small. . As a result, the evaluation value of the search vector does not become wavy as shown in Fig. 2, but becomes flat as shown in Fig. 3, which allows accurate movement that correctly indicates the movement of the subject on the image. It becomes possible to detect vectors.

Based on the above principle, by repeating the operations from the vector evaluation value calculation unit 3 to the block size expansion unit 7 in FIG. 1 until a highly reliable vector is detected or the set number of vector detection stages is reached. , it becomes possible to find vectors that correctly indicate the movement of objects in the supplied images.

(Example) Hereinafter, an example of the present invention will be described in detail with reference to FIGS. 4 and 5. In this embodiment, the blocks are enlarged in two stages, but a configuration in which there are three or more stages is also possible.

The supplied image signal is supplied to the frame memory 9 and also to the subtractor 12 via a line 901.

The supplied image signal is stored in frame memory 9 and supplied to variable delay unit 10 via line 902.

Search vector generation circuit 11 generates a vector according to a determined search range and supplies it to variable delay unit 10 via line 1101.

In the variable delay unit 10, the image signal supplied from the frame memory 9 is delayed according to the vector supplied from the search vector generation circuit 1, and is supplied to the difference unit 12 via the line 1001.

The difference unit 12 calculates a difference value between the supplied image signals and supplies it to the evaluation value calculation circuit 13 via a line 1201.

The evaluation value calculation circuit 13 calculates the evaluation value for each search vector in each divided block of the image from the supplied prediction error signal, for example, the sum of absolute values and the sum of squares of the prediction errors, and sends the result through the line 13o1. and supplies it to the memory 14.

The memory 14 stores evaluation values for each search vector in each divided block of the supplied image, and supplies them to the vector detection circuit 15 via a line 1401.

The vector detection circuit 15 detects one vector with the best evaluation result supplied from the memory 14, or if necessary, detects a vector with the best evaluation result and a plurality of vectors with subsequent evaluation results, and then 1
The signal is supplied to the reliability determination circuit 16 via 501.

The reliability determination circuit 16 determines the reliability of the supplied vector, and if it is determined that the reliability is high, it transmits the vector, and if it is determined that the reliability is low, it transmits the information via the line 1601. The signal is supplied to the motion vector determining circuit 17.

As a method for determining reliability, for example, a method may be used in which several vectors with high evaluation results are selected and if they are close to each other, the reliability is high. Further, depending on the reliability evaluation method, a configuration may be adopted in which necessary signals are obtained from the memory 14, the vector detection circuit 15, or the like.

The motion vector determination circuit 17 records the supplied motion vector detection results, and for blocks for which only unreliable vectors have been found, a motion vector redetection request for that block is sent to the evaluation value calculation circuit via a line 1702. 13 and the search vector generation circuit 11.

Possible methods for enlarging the size of a block include, for example, enlarging it by a fixed number of pixels in the upper, lower, left, and right directions, or enlarging it by a fixed number of pixels in each of the horizontal and vertical directions. Fourth
In Figure (a), blocks shaded with diagonal lines indicate blocks in which highly reliable vectors were found in the first stage of motion vector detection. In addition, FIG. 6B shows how the block size of a block for which a highly reliable vector was not found in the first stage of motion vector detection has increased. The second stage of motion vector detection is performed on this enlarged block. However, at this time, the block to which the detected motion vector is applied is the original size block.

In the search vector generation circuit 11, a search vector is generated in response to a vector re-detection request from the motion vector determination circuit 17, and then similarly to the first stage described above, the search vector is generated by the frame memory 9, variable delay device 10, and difference device 12. , calculates an interframe prediction error signal, and supplies it to the evaluation value calculation circuit 13.

In the evaluation value calculation circuit 13, the motion vector determination circuit 1
In accordance with the vector redetection request supplied from 7, the size of the specified block is expanded in a prescribed manner,
The evaluation value for each search vector of the image is calculated from the prediction error signal supplied to the block in the same manner as in the first stage, and the result is supplied to the memory 14 via the line 1301.

Thereafter, in the memory 14, the vector detection circuit 15, and the reliability determination circuit 16, motion vector detection is performed on the enlarged block as in the case of the first stage, and the detected motion vector or Accordingly, the reliability determination result is sent to the motion vector determination circuit 17 via a line 1601.
is supplied to

In the motion vector determination circuit 17, when the final number of detection stages, in this embodiment the second stage, is completed, the motion vector obtained by the above method is outputted as the final detected motion vector via a line 1701. .

For blocks for which a highly reliable motion vector was not detected in the second stage of motion vector detection, for example, a zero vector may be output, or information indicating that no highly reliable vector was found may be output.

(Effects of the Invention) When detecting a motion vector of an input image signal, it becomes possible to detect a highly reliable motion vector.

[Brief explanation of the drawing]

1, 4, and 5 are diagrams for explaining the present invention in detail, Figures 2 and 3 are diagrams for explaining the invention in detail, and Figure 6 is a diagram for explaining the motion vector in detail. It is a figure for explaining. In the figure, 1... Image input unit, 2... Block size initial setting unit, 3... Vector evaluation value calculation unit, 4... Motion vector detection unit, 5... Reliability determination unit, 6 . . . Vector detection stage number determination section, 7 . . . Block size expansion setting section, 8
...Detected motion vector output unit, 9...Frame memory, 10...Variable delay device, 11...Search vector generation circuit, 12...Differentiator, 13...Evaluation value calculation circuit,
14...Memory, 15...Vector detection circuit, 16
. . . Reliability determination circuit; 17 . . . Motion vector determination circuit.

Claims (1)

[Claims] When detecting a motion vector from a supplied image signal, it has a step of determining the reliability of the detected motion vector, divides the image into blocks of a predetermined size, and performs motion vector detection for each block. For blocks whose detected vectors are not determined to be highly reliable in the reliability determination step, the block size is expanded according to a predetermined method, and the expanded block is subjected to motion vector detection and detection again. The reliability of the detected vectors is determined, and if a highly reliable vector is detected, that vector is output as a detected motion vector in a block of the initially set size, and if a highly reliable vector is not detected. If the detected vector is detected, the motion vector detection method further comprises repeatedly performing a series of operations of enlarging the block size, detecting the motion vector, and determining the reliability of the detected vector.