JPS61184071A - Dynamic vector detection system of animation picture signal - Google Patents

Abstract

PURPOSE:To improve the accuracy of dynamic vector to be detected by executing dynamic vector detection after a forecast vector with high accuracy utilizing acceleration is utilized in detecting the dynamic vector of an animation picture signal. CONSTITUTION:An acceleration 16 is obtained by a multiplier 6 and an adder 7. The acceleration 16 is subtracted by an adder 8 from an address signal 17 representing the present block position detecting the dynamic vector, inputted to a field memory 10 stored with a field signal of one preceding frame together with the similar result 19 in the direction (y), the corresponding picture signal is outputted by a line 20 in the unit of block and inputted to a dynamic vector detection section 12. A picture signal of a block location detecting the dynamic vector from the present field is read from a field memory 11 and inputted to a detection section 12 from a line 22. The detected dynamic vector is stored in a memory 3 from a line 23 for the similar processing of the next field.

Description

DETAILED DESCRIPTION OF THE INVENTION (Field of Industrial Application) The present invention relates to a motion vector detection method for moving image signals, and in particular to a motion vector detection method that provides an initial deviation vector to improve detection efficiency and accuracy. Prior Art) Television video signals are composed of I frames per second in Japan, but interframe coding methods that utilize correlation between frames to achieve high efficiency coding, In a frame interpolation method for converting to a different frame rate, the basic technology is a technique for accurately detecting a motion vector of an image signal between frames.

Conventionally, in order to detect this motion vector, (1
) A method that utilizes signal correlation between frames, (2) A sequential method based on the relationship between inter-frame signal differences and intra-frame signal differences (“Image motion amount detection method” Patent application 1987-13)
501) etc. - These methods detect motion vectors using a plurality of image signals as a unit, but they are basically based on reading out image signals that have been digitized and stored in a frame memory in units of blocks.

As shown in Figure 2, method (1) is to move block A (hatched) for which motion vector detection is to be performed in the current 7 frames to the same position as the center of block A in the previous frame. Evaluate the difference between the signals in all blocks within the range hole where the center of the image signal block is located, and calculate the area R that gives the minimum value.
The difference between the coordinates within and the center point is used as a motion vector.

In addition, method (2) uses the gradient method, which is a well-known technique, repeatedly for each block of the image signal, and when repeating, cuts out a block in the previous frame that is shifted in the opposite direction by the detected motion vector obtained last time. The gradient method is applied again to the block, and is repeated until the amount of detected motion vectors becomes less than a certain value, and the sum of the detected motion vectors obtained up to that point is used as the motion vector for that block.

(Problem to be Solved by the Invention) However, when applying these methods to general moving images such as television signals, method (1) has problems with the amount of motion that can be detected due to hardware scale issues. It is necessary to limit the amount of motion to a certain range, in which case the detection ability becomes zero for the amount of motion that exceeds that range, and when performing interframe coding, the amount of coded information increases rapidly, and In the case of inter-frame interpolation, the quality of the interpolated frame is significantly degraded. In addition, it has become clear that method (2) has performance limitations due to the number of repetitions of detection processing for large amounts of motion, and the detection accuracy deteriorates to some extent. ing.

The present invention has been made in view of the above-mentioned drawbacks of the prior art, and it is an object of the present invention to provide a method for improving the detection accuracy of the conventional motion vector detection method by adding a small amount of hardware.

(Means for Solving the Problem) Therefore, the present invention focuses on the fact that motion vectors of moving image signals change continuously in accordance with changes in frames, and utilizes the acceleration of motion vectors. Then, a motion vector at a block position where a motion vector is to be detected is predicted, and a conventional motion vector detection method is applied focusing on a block at a position shifted in advance by this predicted vector in the previous frame.

(Function) With this means, the detection ability and accuracy are not degraded even for motion amounts exceeding the detection range in conventional motion vector detection methods, and the search range or number of times for detection can be reduced. Can be done.

(Principle of the invention) Hereinafter, the present invention will be explained in detail using the drawings.

A moving image signal is composed of consecutive frames, and a motion vector is detected between two consecutive frames. On the other hand, in televisions and the like, an l frame is further composed of two fields, so motion vector detection is performed between even fields or between odd fields. Since detection is performed anew for each field, the motion vectors detected in the previous field and the field before the previous can be used to detect the motion vector in the current field.

In this case, a method using only information from the previous field may be considered, but in actual television signals and the like, the image signal includes a large amount of acceleration. Therefore, if a vector obtained by the following equation is used as a predicted vector of a motion vector in the current field, the prediction error will be smaller.

For a block whose motion vector is to be detected, the average value of the motion vectors of nine blocks including this block is used. In this case, the number of blocks does not necessarily need to be nine; the range may be further expanded, or for reasons of calculation speed, it may be eight, and the division may be executed by shifting a shift register.

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

FIG. 1 is a diagram showing an embodiment of the present invention. In Figure 1, l
and 2 are memories for storing all the motion vector detection results in each block of the previous field and the previous frame (previous field), and 3 is a memory for storing all the motion vector detection results in each block of the current field. memory; 4.5 is a circuit for calculating the average value of detected vectors in a plurality of blocks as shown in FIG. 2 for the current block position for which motion vector detection is to be performed; 6 is a multiplier; 7.8. 9 is an adder, 10.11 is a field memory, and 12 is a motion vector detection circuit.

Notes 1, 11, 2, 3, average value calculation circuits 4, 5, multiplier 6, and adder 7 are prepared for the X direction and Y direction, since the motion vector to be detected is two-dimensional data. There is a need to. A surrounded by a broken line in Figure 3
, B are circuits for the X direction and the Y direction, respectively, and since they can be realized with exactly the same configuration, only the operation on the A side will be described hereafter.

The contents of memory 1.2.3 are updated sequentially as the field number advances, moving the contents to the next memory as shown by line 13 (average value calculation circuit 4.5 is an adder and a divider). Alternatively, it is configured with a shift register, and for the current block position where motion vector detection is to be performed, the motion vectors detected in multiple blocks as shown in FIG. 2 are sequentially read out from memories 1 and 2, and the average value thereof is determined. .Output signal 14 of average value calculation circuits 4 and 5.
15 are respectively ■(-'L V(-2)
corresponds to the ingredient.

According to formula (1), multiplier 6 and adder 7 calculate formula (11
The left side No.), that is, the acceleration 16 is determined. acceleration 16
is subtracted by the adder 8 from the address signal 17 indicating the current block position for which a motion vector is to be detected, and the same result 19 in the X direction is subtracted from the field memo 1, 1 where the field signal of the previous frame is stored.
10, the corresponding image signal is outputted block by block, and is inputted to the motion vector detection section 12. At the same time, the image signal of the block position for which motion vector detection is to be performed from the current field is stored in the field memory 11.
The motion vector detection section 12 is read out by the motion vector detector 22 and inputted to the motion vector detection section 12 by the motion vector detection section 22 . The detected motion vector is stored in the memory 3 for similar processing on the next field. 21 is the input terminal of the moving image [No. 4] to the field memory. Memo IJI, 2.3 and field memory 10.11 are updated synchronously in units of fields.

(Effects of the Invention) According to the present invention, when detecting a motion vector of a moving image signal,
Motion vector detection is performed after calculating a highly accurate predicted vector using acceleration for the motion vector, so even if the motion vector detection range is small, the motion vector can be detected by sufficiently tracking the actual amount of motion. This makes it possible to reduce the scale of the hardware and improve the accuracy of detected motion vectors.

[Brief explanation of the drawing]

FIG. 1 is a block diagram showing an embodiment of the present invention, FIG. 2 is an explanatory diagram of a conventional motion vector detection method using signal correlation between frames, and FIG. 3 is an explanatory diagram of a block for calculating the average of motion vectors. 1.2,3; memory; 4,5; average value calculation circuit;
6; Multiplier, 7, 8; Adder, 10, 11
; field memory; 12; motion vector detection circuit;

Claims (1)

[Claims] In a motion vector detection method for a moving image signal that detects a motion vector of a block in an image between a previous frame and a current frame, the position of a block in the previous frame is determined by calculating the position of the block in the previous frame by calculating the motion vector of a block in the previous frame. The average value of the motion vectors of multiple blocks ▲ There are mathematical formulas, chemical formulas, tables, etc. ▼ and the average value of the motion vectors of multiple blocks surrounding this value and the block two fields ago ▲ There are mathematical formulas, chemical formulas, tables, etc. ▼ Differences with ▲ There are mathematical formulas, chemical formulas, tables, etc. ▼
A motion vector detection method for a moving image signal that detects a motion vector by shifting the sum determined from ▲There are mathematical formulas, chemical formulas, tables, etc.▼.