JP2519113B2 - Method of transmitting motion vector information, transmitter and receiver thereof - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION [Industrial field of application] The present invention relates to motion vector information necessary for performing motion compensation and motion correction processing in a recording, transmission and display device for processing digital signals which are image signals. TECHNICAL FIELD The present invention relates to a method of transmitting data, a transmitter and a receiver used for the method.

[Prior Art] In high-efficiency coding of moving images, motion-compensated prediction is known in which the previous frame is shifted according to the motion of the image to perform inter-frame prediction.

In order to perform this motion compensation prediction, it is necessary to obtain a motion vector between frames and send the information to the receiving side. 8x8 pixels or 16x16 for motion vector detection
It is common to perform matching in block units of pixels. Transmission of the detected motion vector is performed by variable length coding by utilizing the low occurrence frequency of large motion.

In the conventional example, the smaller the motion vector detection block is, the more complex the motion can be handled. However, since the motion vector information is for each block, the smaller the block, the larger the amount of information to be transmitted. For example, if the size of a block is halved both vertically and horizontally, the amount of information will be quadrupled by simple encoding. Therefore, Japanese Patent Laid-Open No. 1-265684 discloses a technique for transmitting by thinning out blocks.

[Problems to be Solved by the Invention] According to the technique disclosed in Japanese Patent Laid-Open No. 1-265684, interpolation is performed by transmitting motion vectors of peripheral blocks according to a predetermined rule, so that optimum interpolation is always performed. There is no guarantee that it will be done. Especially when there is a large difference between neighboring motion vectors, interpolation of blocks that do not transmit motion vectors may be performed with inappropriate motion vector values due to thinning, so interpolation values far from the original values were used. As a result, it becomes difficult to perform motion compensation on an image with a large amount of motion, resulting in deterioration of image quality and increase of the generated code amount.

Therefore, according to the present invention, the amount of information does not increase when transmitting a motion vector detected in a small block, and a motion vector transmission method and a motion vector transmission used therefor capable of transmitting an image with intense motion without deteriorating the image quality significantly. It is an object of the present invention to provide a receiver and a motion vector receiver.

[Means for Solving the Problem] In order to achieve the above object, according to the present invention, a motion vector is obtained in a unit of a relatively small block, and motion vector information of all blocks is not transmitted but thinned and transmitted. Blocks should be interpolated from around,
As an interpolation method, if the motion vectors of a plurality of adjacent blocks are close to each other, the average value is used to interpolate, and if the motion vectors are significantly different from each other, the motion vector having the closer motion vector is used to perform the interpolation. The transmitting side determines which block is closer to the vector, and transmits the information designating the block to the receiving side.

In the inter-frame predictive coding, motion-compensated prediction is performed using motion vectors that are thinned out and interpolated.

That is, according to the present invention, when transmitting the motion vector information of the image detected for each block on the transmitting side to the receiving side, the detected motion vector information is thinned out in blocks of predetermined intervals and transmitted. For a non-transmission block in which motion vector information is not transmitted, the motion vector of the peripheral block to which the motion vector information is transmitted is used on the receiving side so that the motion vector can be interpolated, A motion vector of the non-transmitted block, the motion vector of the non-transmitted block being detected and transmitted, and the detected motion vector is transmitted. Means for thinning out information at the predetermined interval and motion vector information not transmitted by thinning out at the predetermined interval Lock, in its adjacent blocks, detects whether closer to the motion vector information of any block, the motion vector information transmitter is provided with means for transmitting the information.

Further, according to the present invention, when transmitting the motion vector information of the image detected for each block on the transmitting side to the receiving side, the detected motion vector information is thinned out in blocks of predetermined intervals and transmitted. For a non-transmission block in which motion vector information is not transmitted, the motion vector of the peripheral block to which the motion vector information is transmitted is used on the receiving side so that the motion vector can be interpolated, A receiver used in a method of transmitting motion vector information, which comprises detecting and transmitting information indicating which block motion vector is closest to the motion vector of the non-transmission block,
Means for interpolating the transmitted motion vector information at predetermined intervals, and for blocks transmitted from the transmission side and not transmitting motion vector information by thinning out at the predetermined intervals, of any adjacent blocks Provided is a motion vector information receiver having means for selecting a motion vector of the closer adjacent block as a motion vector to be interpolated in response to information indicating whether the motion vector is close to the motion vector.

[Operation] According to the motion vector information transmission method having the above-described configuration, first, the number of motion vectors is reduced by thinning out on the transmission side, so that the amount of information to be transmitted is significantly reduced. Furthermore,
When the change in the motion vector is small between blocks, the thinned-out vector is smoothly interpolated on the receiving side to obtain a smoothly changing vector.

When the movement is largely different in each area, such as a person moving in front of the background, interpolation is performed at either of the boundaries of the area, and the block is appropriately divided into areas.

Embodiments Embodiments of the present invention will be described below with reference to the drawings. FIG. 1 is a block diagram showing a motion vector information transmitter, a transmission system, and a motion vector information receiver that realize the motion vector information transmission method of the present invention. In the figure, TX indicates the transmitter,
RX indicates a receiver. A transmission system TS is provided between the transmitter TX and the receiver RX. First, in the transmitter TX, the digital image signal input from the image input terminal 10 is first detected by the block motion vector detector 12 in a block of 8 × 8 pixels, similarly to the conventional device, to detect a motion vector (MV). .
The detected motion vector is a motion vector horizontal thinning circuit.
At 14, horizontal blocks are thinned out every other block. The output is further 1 in the vertical direction by the motion vector vertical thinning circuit 16.
It is thinned out every other block and transmitted to the receiver RX. Since every other block is arranged vertically and horizontally, one motion vector is transmitted to 16 × 16 pixels, and the image is 352 × 2.
If there are 40 pixels, the number of vectors per frame is 44 × 30 = 1320, but it becomes 22 × 15 = 330, and the number becomes 1/4.

In the receiver RX, blocks for which motion vector information has not been transmitted are interpolated from upper and lower or left and right blocks. In the interpolation process, the horizontal and vertical directions are reversed from those of the transmitter TX. The motion vector vertical interpolation circuit 18 first performs vertical interpolation, and then the motion vector horizontal interpolation circuit 20 performs horizontal interpolation. The order of horizontal and vertical interpolation may be exchanged between the transmitter TX and the receiver RX, but the method shown in the figure is advantageous from the viewpoint of the capacity of the line block memory of vector information required for vertical processing.

The processing of each block is shown in FIG. Since the horizontal direction is interpolated later, the blocks diagonal to the transmission block are interpolated horizontally using the information of the blocks already interpolated in the vertical direction.

Next, a concrete method of the thinning and interpolation will be described with reference to FIG. FIG. 2 is a block diagram showing a specific configuration of the motion vector vertical thinning circuit 16 and the motion vector vertical interpolation circuit 18 of FIG. This FIG. 2 shows a method of vertical decimation and interpolation, but in the case of horizontal, the line block delay units 32 and 38 are replaced by one block delay units,
There is no difference from the vertical direction except that the judgment described later is performed.

The input motion vector is thinned out by the switch 30 on a block-by-block basis, and that of the transmission block is input to the 1-block line delay unit 32 and the vector subtractors 34 and 44. switch
30 is driven by the output signal of the 1/2 frequency divider 28, and this 1/2 frequency divider 28 is a synchronization detection circuit 26 for extracting a synchronization signal synchronized with the motion vector rate from the input signal. By dividing the output synchronization signal of the
The contact is switched for each block. The motion vector of the non-transmitted block is input to the subtractors 44 and 46. The 1-line block delay unit 32 delays the motion vector data by the number of blocks in the horizontal direction in a thinned state. However, since the motion vector data is represented by a binary value of a horizontal motion amount and a vertical motion amount, it becomes a binary memory. The output of the one-block delay unit 32 is transmitted as motion vector information as in the conventional device.

The vector subtractor 34 takes a horizontal and vertical difference between the transmission blocks to obtain a difference vector. The distance vector (RMS) 36 determines the square root of the sum of squares of the horizontal and vertical differences. This corresponds to the Euclidean distance of two vectors, and when this is larger than a predetermined value TH (for example, 3 pixels), the comparator 37 sends a signal of "H" level and controls the switch 53 described later to interpolate mode. Transmit information. However, in horizontal decimation, the blocks deleted in the subsequent vertical decimation cannot detect the correct distance at the receiver RX, so this decision is not made and the decision on the upper line is used instead.

For the vector of the non-transmission block, the distance between the vector of the transmission block obtained by the vector subtractor 44 and the distance calculator 48 and the vector of the previous transmission block obtained by the vector subtractor 46 and the distance calculator 50 is calculated. To be done. That is, the subtractor 44 and
It is determined which of the 46 output signals has a smaller value, that is, which is closer. In this example, the comparator 52 outputs "1" when the output signal of the subtractor 44 is smaller than the output signal of the subtractor 46, and outputs "0" in the opposite case. This signal is transmitted to the receiver RX as the interpolation mode information when the distance of the block transmitted via the switch 53 is large.

Next, the interpolation in the receiver RX will be described. First, the transmitted motion vector is input to the line block delay unit 38, the subtractor 40, and the vector adder 60. The distance of the transmission vector is obtained by the subtractor 40 and the distance calculator (RMS) 42 in the same manner as in the case of thinning, and it is determined whether the interpolation mode information has been transmitted. As a result, the switch 56 is controlled and the following two types of interpolation methods are switched.

When the difference in motion of the transmission block is large, the switch 53 is turned on by the output signal of the comparator 37 and the interpolation mode information is transmitted to the receiver RX. One of the corresponding motion vectors is selected and becomes an interpolation vector, which is output via the switches 56 and 58. On the other hand, when the movement difference between the transmission blocks is small, the switch 53 is off, the interpolation mode information is not transmitted, and the interpolation is performed on the average of the vertical and horizontal directions. An example of this average interpolation is shown in FIG. A block between a block of 4 pixels on the top, a motion of 2 pixels to the right (MV1) and a block of 2 pixels on the top, and a motion of 4 pixels to the right (MV3) is 3 pixels on the top and 3 pixels on the right ( MV2). This calculation is performed by the vector adder 60, and the calculation result is output via the switches 56 and 58. The comparator 62 in the receiver RX has the same configuration as the comparator 37 in the transmitter TX, and operates similarly. Further, the synchronization detection circuit 64 is similar to the synchronization detection circuit 26 of the transmitter TX, but its output signal is directly used as a control signal of the switch 58 without passing through a frequency divider, and an interpolation operation is performed.

With the above processing, it is possible to perform appropriate interpolation even at the boundary between different motion areas. An example thereof is shown in FIG. It is assumed that the movements in the shaded area and the other areas in the figure are significantly different.

a, b, c, d are transmission blocks, and the motion vector remains unchanged. In the non-transmission block, 2 and 4 are vertically interpolated. Since a and c and b and d are the same moving area, the difference between the motion vectors is small and the average value is interpolated.

1, 3 and 5 are interpolations from the horizontal direction, but a, b and c
And d are different moving regions, and the motion vectors are greatly different. Therefore, the interpolation mode information is 1, 2, 3
All are transmitted, and 1 is a, 3 is 4, 5 is d from the area of the moving area in the block. Therefore, a, 1,
The moving area is divided into 2, c and b, 3, 4, 5, and d, and appropriate interpolation processing is performed. FIG. 5 is a diagram showing an example of vertical (vertical) compensation and horizontal (horizontal) interpolation in different moving areas.

[Effects of the Invention] As is clear from the above description, according to the motion vector transmission method of the present invention, the transmitting side thins out the blocks to transmit the motion vector, and the thinning out blocks the blocks that are not transmitted to the receiving side. In order to be able to adaptively interpolate from, the motion vector of the block in the peripheral block of the non-transmitted block is detected, and the amount of the motion vector to be transmitted is reduced by transmitting the block information. Therefore, the amount of information to be transmitted is significantly reduced and the deterioration of image quality is small. Also, when the change of the motion vector is small between blocks, the thinned-out vector obtains a smoothly changing vector by the average interpolation from the surroundings, and the block distortion becomes smaller than transmitting all the motion vectors. In addition, when the motion is largely different in each area such as a person moving in front of the background, interpolation is performed at either of the boundaries of the area, and the block is appropriately divided into areas. As a result, the amount of data can be further reduced by motion compensation coding or the like.

[Brief description of drawings]

FIG. 1 is a block diagram showing a configuration of a system that realizes a motion vector information transmission method of the present invention, and FIG. 2 is a block diagram showing configurations of a vertical thinning circuit and a vertical interpolation circuit of FIG.
FIG. 3 is a diagram showing an example of forming an average value interpolation vector, FIG. 4 is a diagram showing a block state in the present invention, and FIG. 5 is a diagram showing an example of interpolation in different moving regions. 10 ... Image signal input, 12 ... Block motion vector detection circuit, 14 ... Motion vector horizontal thinning circuit, 16 ... Motion vector vertical thinning circuit, 18 ... Motion vector vertical interpolation circuit, 20 ... Motion vector horizontal interpolation circuit Circuit, 22 …… Motion vector output, 30, 53, 54, 56, 58 …… Switch, 32, 38
...... Line block delay device, 34, 40, 44, 46 …… Vector subtractor, 36, 42, 48, 50 …… Distance calculator, 37, 52, 62
…… Comparator, 60 …… Vector adder, TH …… Predetermined value, TS
...... Transmission system, TX …… Transmitter, RX …… Receiver.

Claims (4)

(57) [Claims] 1. When transmitting motion vector information of an image detected for each block on the transmission side to a reception side, thinning out the detected motion vector information in blocks at predetermined intervals and transmitting the motion vector information. For the non-transmission block in which the vector information is not transmitted, which one of the peripheral blocks is included so that the motion vector can be interpolated using the motion vector of the peripheral block in which the motion vector information is transmitted on the receiving side. A motion vector of the block, a transmitter used in a method of transmitting motion vector information, which comprises detecting and transmitting information indicating whether the motion vector of the non-transmission block is closest, Means for thinning out at the predetermined interval and blocks for which motion vector information is not transmitted due to thinning out at the predetermined interval. Te in its adjacent blocks, detects whether closer to the motion vector information of any block, the motion vector information transmitter having means for transmitting the information. 2. A means for detecting whether or not a block-to-block distance of motion vector information to be transmitted after being thinned out at the predetermined interval is a predetermined value or more, and only when the distance is the predetermined value or more. 2. The motion vector information transmitter according to claim 1, further comprising means for sending information indicating whether the motion vector is close to the motion vector of the adjacent block of. 3. A step of thinning out the detected motion vector information in blocks at predetermined intervals when transmitting the motion vector information of the image detected for each block on the transmitting side to the receiving side, and a motion. For the non-transmission block in which the vector information is not transmitted, which one of the peripheral blocks is included so that the motion vector can be interpolated using the motion vector of the peripheral block in which the motion vector information is transmitted on the receiving side. A receiver used in a method of transmitting motion vector information, which comprises detecting and transmitting information indicating whether a motion vector of a block is closest to a motion vector of the non-transmitted block, wherein the transmitted motion vector information Is interpolated at a predetermined interval, and information transmitted from the transmitting side is thinned out at the predetermined interval to obtain a motion vector. Motion vector information having means for selecting a motion vector of the closer adjacent block as a motion vector to be interpolated in response to information indicating which of the adjacent blocks the motion vector of which is not transmitted Receiving machine. 4. A means for detecting whether or not a block-to-block distance of motion vector information thinned and transmitted at a predetermined interval is a predetermined value or more, and any one of the adjoining means only when the distance is the predetermined value or more. 4. The motion vector information receiver according to claim 3, further comprising means for selecting a corresponding motion vector using information indicating whether the motion vector is close to the motion vector of the block.