JPH07107484A - Picture transmitter - Google Patents

Abstract

PURPOSE:To make a search effective and to shorten the processing time by providing a memory to be outputted as the former moving vector and limiting the search range in a moving compensator based on the former moving vector. CONSTITUTION:A quantization number 53 to be outputted for a reverse quantization device 5 becomes a reproduction differential block signal 55 and is outputted to an adder 6. The block signal 55 is added with an estimate signal 58 in the adder 6 and is stored in a frame memory 7 as a reproduction block signal 56. The stored signal 56 is outputted for a moving compensator 8 as the former frame signal 57 and is used for the next frame processing. The moving compensator 8 compares the signal 57 with the present block signal 51 for each block and selects the one with less distortion. The selected block signal is outputted as an estimate signal 58 and the difference in position between the present block signal 51 and the selected one is outputted as a moving vector 59.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION The present invention relates to an image communication system,
In particular, the present invention relates to a motion compensation type image transmission device for compressing, encoding and transmitting an image signal in a video conference, a video telephone and the like.

[0002]

2. Description of the Related Art In image communication such as a video conference and a video phone, when an image data is transmitted without being compressed,
Since the image information is enormous, a transmission line having a high communication speed is required, which makes it difficult to realize the transmission line and increases the transmission cost. Therefore, compress the amount of information of the image to be transmitted,
Less transmitted data is done. The motion compensation method, which is one of these compression methods, has a high degree of compression and is excellent.

The motion compensation method is a method that is conceived to prevent the coding efficiency of a moving image from being lowered, and the image supplied at a predetermined frame time interval is used. Each block (frame) is divided into small blocks of a predetermined size, and the motion vector (displacement information) between the frames for each of these small blocks is divided.
Is detected, compensation is performed using these motion vectors, and reproduction is performed.

FIG. 7 is a block diagram showing the structure of an image transmission apparatus using the conventional motion compensation method described in Japanese Patent Laid-Open No. 2-299380. In the figure, reference numeral 1 denotes a signal obtained by converting an image signal obtained by a video camera or the like from an analog signal to a digital signal, and then blocking adjacent pixels on the image for every K pixels (K is an integer of 2 or more). An input signal line 2 to which (hereinafter, referred to as a current block signal 51) is input and a difference between the current block signal 51 from the input signal line 1 and a prediction signal 58 output from the motion compensator 8 is calculated and output. The subtractor 3 encodes the difference signal sequence (difference block signal 52) output from the subtracter 2 by comparing it with a predetermined quantization table and selecting the optimum quantization number 53. Quantization number 53 output from the quantizer 3
Is encoded with a motion vector 59 into a variable-length bit signal and output, and a variable-length encoder 5 reproduces a difference signal sequence from a quantization number 53 output from the quantizer 3 using a predetermined quantization table. Then, the inverse quantizer that outputs the reproduced difference block signal 55, and the adder 6 adds the reproduced difference block signal 55 that the inverse quantizer 5 outputs and the prediction signal 58 and outputs the reproduced block signal 56 to the frame memory 7. An adder 7, a frame memory for storing image data (frame) based on the reproduction block signal 56, 8 a motion compensator for obtaining a prediction signal 58 and a motion vector 59 from the current block signal 51 and the previous frame signal 57, 9 Is an output signal line for outputting the output (transmission data 54) of the variable length encoder 4 to the outside.

Next, the operation will be described. The image input signal is converted from an analog signal to a digital signal, and is then divided into blocks by collecting adjacent pixels on a single image (frame) for each predetermined number (K).
The current block signal 51 of the dimension is input from the input signal line 1.

The current block signal 51 is input to the subtractor 2 and the motion compensator 8. In the motion compensator 8, the prediction signal 58 is calculated from the current block signal 51 and the previous frame signal 57 stored in the frame memory 7.
To generate. This prediction signal 58 is the current block signal 5
The block data having the smallest distortion with 1 is selected from the frame memory 7. At the same time, a motion vector 59 is generated and output to the variable length coding circuit 4 (detailed operation of the motion compensator 8 will be described later).

Next, the subtractor 2 calculates the difference between the current block signal 51 and the prediction signal 58 to generate a difference block signal 52, which is output to the coding quantizer 3. The difference block signal 52 is quantized by being compared with a predetermined quantization table in the quantizer 3 and selecting the optimum quantization number 53. The quantization number 53 is output to the variable length encoder 4 and the inverse quantizer 5.

In the variable length coding circuit 4, the quantization number 53
Together with the motion vector 59 is converted into a variable length code and multiplexed to form transmission data 54 which is output via the output signal line 9. The encoded multiplexed signal 54 has a variable data length, and if it is a still image, the data size is small. In the case of a moving image, the larger the motion, the larger the data size.

On the other hand, the quantization number 53 output to the inverse quantizer 5 is inversely quantized based on a predetermined quantization table similar to the case of the quantizer 3, and becomes a reproduction difference block signal 55 and an adder. 6 is output. Then, the adder 6 performs addition with the prediction signal 58 to obtain the reproduction block signal 5
6 is stored in the frame memory 7. The stored reproduced block signal 56 is used by the motion compensator 8 when processing the next frame. By repeating the above operation for each frame, a moving image is transmitted.

The operation of the motion compensator 8 will be described in detail. The motion compensator 8 uses the previous frame signal 57 stored in the frame memory 7 for the current block signal 5 for each block.
Compared with 1, select the block with the least distortion (approximate block). Then, the selected block signal is output as the prediction signal 58, and the position difference (displacement information) between the current block signal 51 and the selected block signal is output as the motion vector 59.

The above operation will be described with reference to FIG. In the figure, 36 meshes (X00-X5) separated by dotted lines
5) shows the previous frame signal 57 stored in the frame memory 7, and each mesh thereof shows one pixel of the image signal. On the other hand, when the current block signal 51 is superimposed on FIG. 6, it is assumed that the block is a 2 mesh × 2 mesh block (X22, X23, X32, X33) indicated by a solid line P. Then, the coordinates of the origin (X22) are set to A (0,0) (in the figure, the upper left of the pixel (mesh) is set as the point indicating the coordinates of the pixel). In order to compare the current block signal 51 and the block of the previous frame signal 57, in the previous frame signal 57 of 6 mesh × 6 mesh in FIG. 6, while moving the block of the current block signal 51 within a range of ± 2, The blocks that give the least distortion may be searched sequentially. The blocks to be moved are called transition blocks (Q and R in FIG. 6).

By the way, the distortion calculation is performed by obtaining the sum of absolute differences between the corresponding meshes in the block, or by obtaining the Euclidean distance. here,
Current block signal 51 and block Q (X11, X12, X21, X22)
This block is selected as the prediction signal 58 when the distortion between and is the smallest, that is, the sum of absolute differences is the smallest. The motion vector is (-1,1). Similarly, when the block R (X44, X45, X54, X55) is selected, its motion vector is (2, -2).

Here, in order to search the entire previous frame signal 57 of FIG. 6, since the origin of the block is the mesh X22, the horizontal and vertical directions are -2, -1, 0, + with respect to the origin, respectively.
Distortion must be calculated for each block that has transitioned by 1, +2. Therefore, in the case of FIG.
The origin must be changed and calculated 5 × 5 = 25 times. Therefore, as the number of meshes increases, the processing time will increase significantly.

[0014]

In the conventional image transmission apparatus, as described above, the prediction signal 58 in the motion compensator 8 is used.
Also, in order to obtain the motion vector 59, the entire frame signal 57 had to be searched and calculated last time.

The present invention has been made to solve the above problems, and an object thereof is to provide an image transmission apparatus capable of shortening the processing time by limiting the search range. .

[0016]

According to a first aspect of the present invention, an image transmission apparatus includes a memory that stores a motion vector and outputs the motion vector as a previous motion vector, and a motion compensator performs a predetermined search based on the previous motion vector. A region is defined and the previous frame signal is searched for in the search region.

An image transmitting apparatus according to a second aspect includes a memory for storing a motion vector and outputting it as a previous motion vector, and a bit for setting a predetermined search area determined on the basis of the previous motion vector in the motion compensator. A switch is provided, and the motion compensator is configured to search for the previous frame signal based on the bit switch.

According to a third aspect of the present invention, in the image transmitting apparatus, a distortion allowable range is set in advance in the motion compensator, and when the motion compensator searches for the previous frame signal based on the current block signal, the distortion is the distortion allowable range. The configuration is such that the search is aborted when it falls within the range.

An image transmitting apparatus according to a fourth aspect comprises a transmission path condition monitor for monitoring the condition of the transmission path, and the motion compensator has a distortion allowable range determined based on the output of the transmission path condition monitor. The motion compensator is set and the search is terminated when the distortion when the previous frame signal is searched based on the current block signal falls within the distortion allowable range.

[0020]

According to the first aspect of the invention, the memory stores the motion vector and outputs it as the previous motion vector, and the motion compensator defines a predetermined search area based on the previous motion vector. The previous frame signal is searched.

According to the second aspect of the invention, the memory stores the motion vector and outputs it as the previous motion vector, and the motion compensator sets the predetermined search area determined based on the previous motion vector. The previous frame signal is searched based on.

According to the third aspect of the present invention, a distortion allowable range is set in advance in the motion compensator, and the distortion when the motion compensator searches for the previous frame signal based on the current block signal is within the distortion allowable range. When, the search is terminated.

In the fourth aspect of the present invention, the transmission line condition monitor monitors the condition of the transmission line, and the motion compensator sets a distortion allowable range determined based on the output of the transmission line condition monitor. The motion compensator terminates the search when the distortion when searching for the previous frame signal based on the current block signal falls within the distortion allowable range.

[0024]

EXAMPLES Example 1. FIG. 1 is a block diagram showing the configuration of an image coding apparatus according to an embodiment of the present invention. In the figure, 1 is a signal obtained by converting an image signal obtained by a video camera or the like from an analog signal to a digital signal, and then blocking K pixels (K is an integer of 2 or more) that are close to each other on the image. An input signal line 2 to which (hereinafter, referred to as a current block signal 51) is input, 2 is a difference block for obtaining a difference between the current block signal 51 from the input signal line 1 and the prediction signal 58 output from the motion compensator 8. The subtractor 3 that outputs the signal 52 encodes the differential block signal 52 output from the subtractor 2 based on a predetermined quantization table,
A quantizer 4 that outputs the quantization number 53, a variable length encoder 4 that encodes the quantization number 53 into a variable-length bit signal together with the motion vector 59, and outputs the variable-length bit signal 5 that outputs a difference signal sequence from the quantization number 53. Play and play differential block signal 5
5 is an inverse quantizer, 5 is an inverse quantizer, 6 is an adder for adding the reproduction difference block signal 55 output from the inverse quantizer 5 and the prediction signal 58, and is an output as a reproduction block signal 56 to the frame memory 7. 1 based on block signal 56
A frame memory for storing an image (frame) before the frame, 9 is an output signal line for outputting the output of the variable length encoder 4 to the outside, and these are the same as those in the conventional example.

Reference numeral 8 is a motion compensator for obtaining a prediction signal 58 and a motion vector 59 from the current block signal 51 and the previous frame signal 57. The motion compensator 8 limits the search range based on the previous motion vector 62, and
Based on the transmission status signal 61, a threshold value for aborting the process is set. A memory 10 stores the motion vector 59 detected by the motion compensator 8 and outputs the motion vector 62 to the motion compensator 8 as a previous motion vector 62, and 11 a transmission data amount signal 60 generated by the variable length encoder 4. This is an information amount monitor that monitors and outputs a transmission status signal 61 to the motion compensator 8.

Next, the operation will be described. The current block signal 51 is input to the subtractor 2 and the motion compensator 8. The motion compensator 8 generates a prediction signal 58 from the current block signal 51 and the previous frame signal 57 stored in the frame memory 7. The prediction signal 58 is selected from the frame memory 7 as block data having the smallest distortion with the current block signal 51. At the same time, the motion vector 59 is generated and output to the variable length encoder 4 and stored in the memory 10. This is because when the motion compensator 8 generates the prediction signal 58, a predetermined search process is performed within a certain range to improve efficiency.

Next, the subtractor 2 generates the difference block signal 52 by taking the difference between the current block signal 51 and the prediction signal 58, and outputs it to the encoding quantizer 3. The difference block signal 52 is quantized by being compared with a predetermined quantization table in the quantizer 3 and selecting the optimum quantization number 53. The quantization number 53 is output to the variable length encoder 4 and the inverse quantizer 5.

In the variable length coding circuit 4, the quantization number 53
Together with the motion vector 59 are converted into a variable length code and multiplexed to form a coded multiplexed signal 54 which is output via the output signal line 9. The encoded multiplexed signal 54 has a variable data length, and if it is a still image, the data size is small. In the case of a moving image, the larger the motion, the larger the data size. At the same time, the variable length coding circuit 4 outputs a transmission data amount signal 60 indicating the data amount of the transmission data 54 to the information amount monitor 11. The information amount monitor 11 measures the data amount based on the transmission data amount signal 60, determines whether the transmission amount is larger or smaller than the transmission amount, or whether or not there is a margin in the transmission line. Is output to the motion compensator 8. For example, the variable length encoder 4 uses the data amount (b
(it, byte, or word unit) is output as a transmission data amount signal 60, and the information monitor 11 compares the total amount of this data with the transmission capacity of the output signal line 9 (transmission line) which is known in advance. . If the total amount of data exceeds the transmission capacity or if there is only a slight margin, it is determined that there is no margin in the transmission path, and a signal indicating that is output to the motion compensator 8 as the transmission status signal 61. On the contrary, if the transmission capacity has a sufficient margin with respect to the total amount of data, a signal indicating that the transmission path has a margin is output. The transmission status signal 61 is used to determine how much distortion the search should be terminated as described later.

On the other hand, the quantization number 53 output to the inverse quantizer 5 is inversely quantized to be a reproduced difference block signal 55 and output to the adder 6. The reproduction difference block signal 55 is added to the prediction signal 58 in the adder 6 and stored in the frame memory 7 as the reproduction block signal 56. The stored reproduction block signal 56 is output to the motion compensator 8 as the previous frame signal 57 and is used in the processing of the next frame. By repeating the above operation for each frame, a moving image is transmitted.

Here, the operation of the motion compensator 8 will be described in detail. The motion compensator 8 uses the previous frame signal 57 stored in the frame memory 7 for the current block signal 5 for each block.
Compared with 1, select the one with the least distortion (the one that approximates). Then, the selected block signal is output as the prediction signal 58, and the position difference (displacement information) between the current block signal 51 and the selected block signal is output as the motion vector 59.

By the way, the conventional motion compensator 8 calculates the distortions for all the defined points, finds the minimum distortion among them, and outputs the block. However, generally, in consecutive image frames, images and blocks at the same position have strong correlation and are close to each other. The motion vector also shows a value approximated to the previous frame or neighboring blocks. Therefore,
The block that gives the minimum distortion is almost in the vicinity of the previous motion vector, and it does not change much in terms of accuracy (distortion magnitude) compared to searching the entire range with a lot of processing time. . That is, the processing time when searching at all points is naturally longer than the processing time when the search range is limited / thinned, but the search accuracy (whether distortion is minimum) is not necessarily proportional to the processing time. Therefore, by limiting the search range, it is possible to improve the processing efficiency without lowering the accuracy.

On the other hand, the amount of data transmitted to the transmission line changes depending on the coding situation and block based on the nature of the image to be transmitted. On the other hand, the amount of data (transmission capacity) that can be transmitted through the transmission path is constant. Therefore, the data amount of the motion-compensated image transmission device changes for each image to be transmitted, and the data amount increases when the motion is vigorous as compared with the case of the still image. In a fast-moving image, it will be dropped because it is compressed, or it will take a long time for calculation processing.
The sound and the image may not match (because the sound data is almost constant). That is, when the transmission data amount is larger than the transmission line capacity, the quantized image data is kept waiting. However, in the opposite case, the image data is not kept waiting. Therefore, monitor the status of the transmission path, and if the transmission path has a margin (there is little transmission data), allow a certain amount of reduction in search accuracy to speed up the processing time, improve the data transmission speed, and improve the number of images. On the contrary, when there is no margin in the transmission path (there is a large amount of transmission data), a large amount of search processing is performed during the time waiting for the transmission of image data, improving search accuracy and ensuring high image quality. To do so. By performing the processing as described above, the dead time in the processing of the image transmission device is eliminated, and the performance of the entire device is improved.

Specifically, the processing speed and processing accuracy can be improved by performing the following processing. (1) The motion vector of the previous block in the previous frame or the current frame is stored in the memory, and the range of the search performed this time is determined based on this previous motion vector. (2) In the previous frame search, select a block whose distortion is less than or equal to the threshold value and do not search any more. (3) The search range of (1) and the threshold value of (2) are not set to constant ranges and values, but are changed according to the status of the transmission path.

Next, the operation of the motion compensator 8 in this embodiment will be described with reference to the flowchart of FIG. FIG. 2 shows the processing for one frame.

Step 101 (Determination of Search Range / Threshold) First, the search range and threshold are determined based on the transmission status signal 61 and the previous motion vector 62. That is, if there is a margin in the transmission path based on the previous motion vector 62, the search range is narrowed and the threshold value is increased to speed up the processing. Conversely, if the transmission path is full and there is a waiting time, the search is performed. The range is widened and the threshold value is lowered to process the data over time to improve the search accuracy.

An example of the search range will be described with reference to FIG. FIG. 5 is a diagram showing search coordinates, which are similar to the coordinates of points A, B, and C shown in FIG. In FIG. 5, for example, the previous motion vector 62 (point) is P
It is assumed that it is 1 (1, 1). Since the correlation between frames is high, it is estimated that the motion vector to be obtained from this point is also in this vicinity. Therefore, it is only necessary to search for a point inside the P1 (1, 1) with a constant radius. Then, this radius is determined based on the transmission status signal 61.

Since it is necessary to shorten the processing time if there is a margin in the transmission path, the radius is reduced and the search range is narrowed (9 points) as in the range A1 in FIG. On the other hand, if there is no margin in the transmission path, the processing time may be lengthened, so that the radius is increased as shown in the range A2 (part of which is shown) in FIG.
Widen the search range (25 points). Further, when it is determined that the motion vector to be obtained is not in the center due to changes in the past motion vector, etc., P1 (1, 1) is not set as the center of the range, and the range A3 in FIG. You may come to.

The threshold value is based on the transmission status signal 61, and the capacity of the transmission line and the amount of data to be transmitted are harmonized.
It is set to an appropriate value so that no dead time is generated. If the threshold is small, the search will continue until the distortion becomes smaller than that, so the processing time will be longer. On the contrary, if the threshold is large, the processing will be discontinued for a relatively large distortion and the processing time will be short (however, Falls). That is, the threshold value is variable, and increases if the transmission path has a margin, and decreases if the transmission path has no margin.

Step 102 (Create Bit Switch) Based on the search range obtained in Step 101, a bit switch is created. The bit switch is a switch that controls whether or not the search process is performed in the motion compensator 8. For example, as shown in FIG. 3, it is set for each pixel (search = 1 / no search = 0). The motion compensator 8 searches based on this bit switch. The bit switches in the figure are expanded on the memory, and the vertical axis shows the memory address and the horizontal axis shows the bit at one address. Each bit switch in FIG. 3 corresponds to one pixel in FIG. 5, and the area A1, A2, A in FIG.
3 corresponds to the areas A1, A2, and A3 of the bit switch in FIG. 3, respectively. The block search range is shown in FIG. 4 as Y in FIG.
A setting example of the bit switch in the case of limiting to the range of the region 4 (fourth quadrant) is shown.

Step 103 (switch determination) Next, based on the bit switch set in step 102, it is determined whether or not to search. That is, when the transition block is transitioned and a new search point is to be processed, it is checked whether the bit switch is 1 or 0. If the bit switch is 1 (ON), proceed to step 104 (distortion calculation) and the other. If the switch is 0 (OFF), the process proceeds to step 106 (shift to the next bit switch).

Step 104 (distortion calculation) The distortion (sum of absolute differences, Euclidean distance, etc.) between the transition block and the corresponding previous frame signal 57 is calculated. That is, the absolute value of the difference from the pixel of the previous frame signal 57 corresponding to the pixel of the transition block is taken, and the sum thereof is obtained.

Step 105 (determination of threshold value) The distortion calculated in step 104 is compared with the threshold value to determine whether or not the search for the current block signal 51 is completed. That is, if the distortion calculated in step 104 is smaller than the threshold value determined in step 101, the transition block is set as the block of the previous frame that gives the minimum distortion, and the subsequent processing is terminated.

Step 106 (shift to the next bit switch) When it is determined in step 103 that the bit switch is OFF, the position of the bit switch is moved in order to process the next search point. For example, in FIG. 5, (0,
0) position (address + in the bit switch of FIG. 3)
2, the position of BIT2) is finished, and when the position of (1, 0) in FIG. 5 is searched next, the position on the bit switch is moved to BIT1 with the address unchanged. In step 105 (threshold value judgment),
Even if the abort processing is not performed, the processing of step 106 is performed. The above processing is performed for each block.

The structure of this embodiment has the following effects. First, since the search range is limited and searched based on the previous motion vector, efficient search is possible and processing time can be shortened. For example, when searching the area A1 in FIG. 5, it is sufficient to search for 9 points, which is 9/25 of the processing time when searching for the entire 25 points in FIG. Secondly, since the termination of the processing is determined based on the threshold value, the efficient search can be performed, the processing can be performed in a short time, and the performance is improved. Thirdly, by adjusting the search range and the threshold value according to the transmission situation, efficient processing can be performed without degrading accuracy. That is, when the amount of data is large and the transmission time is long, it is possible to perform a highly accurate search by effectively utilizing the waiting time until the data is transmitted and performing the search process during this time. Fourthly, since the determination of the search is made by a bit switch, the search range can be controlled with a simple configuration.

Example 2. Although the search range and the threshold value are adjusted based on the information amount monitor 11 in the above embodiment, the search range and the threshold value may be set in advance and fixed. According to such a configuration, since the information amount monitor 11 is not required, the configuration is simplified and the processing speed can be improved.

Example 3. Further, although the threshold value is determined in addition to the limitation of the search range in the first embodiment, only the determination of the search range or the threshold value may be processed. For example, when only the search range is limited, step 105 (determination of threshold value) in the flowchart of FIG. 2 becomes unnecessary. Also,
If only the threshold value is used for determination, step 1 in the figure
02 (bit switch creation) and step 103 (switch determination) are unnecessary. With such a configuration, the number of processing steps is reduced, and the processing speed can be improved.

[0047]

According to the inventions of claim 1 and claim 2,
Since the search range in the motion compensator is limited based on the previous motion vector, efficient search is possible and the processing time can be shortened.

According to the third aspect of the present invention, since the search processing of the motion compensator is terminated based on the threshold value, an efficient search is possible and the processing time can be shortened.

According to the fourth aspect of the present invention, the threshold value is made variable according to the transmission status, and the search processing is terminated based on this threshold value, which enables efficient search and Appropriate processing can be performed according to the transmission status.

[Brief description of drawings]

FIG. 1 is a diagram showing a configuration of an image transmission device of the present invention.

FIG. 2 is a flowchart for explaining the operation of the motion compensator of the image transmission device of the present invention.

FIG. 3 is a diagram showing a configuration of a bit switch used in the motion compensator of the image transmission device of the present invention.

FIG. 4 is a diagram showing a specific example of bit switches used in the motion compensator of the image transmission device of the present invention.

FIG. 5 is a diagram for explaining the operation of the motion compensator of the image transmission device of the present invention.

FIG. 6 is a diagram for explaining a search operation in the motion compensator.

FIG. 7 is a diagram showing a configuration of a conventional image transmission device.

[Explanation of symbols]

 1 Input signal line 2 Subtractor 3 Quantizer 4 Variable length encoder 5 Inverse quantizer 6 Adder 7 Frame memory 8 Motion compensator 9 Output signal line 10 Memory 11 Information amount monitor

Claims (4)

[Claims] 1. A motion vector is predicted by searching a previous frame signal stored in a frame memory on the basis of a current block signal in which an image signal is grouped into a plurality of blocks for each pixel and determining a block with less distortion. In a picture transmission apparatus that includes a motion compensator for generating a signal, quantizes the difference between the current block signal and the predicted signal, and quantizes and encodes the motion vector together with the motion vector for transmission. In addition to having a memory for outputting as a motion vector, the motion compensator is configured to determine a predetermined search area based on the previous motion vector and search the previous frame signal in the search area. Image transmission device. 2. A motion vector is predicted by searching a previous frame signal stored in a frame memory on the basis of a current block signal in which an image signal is grouped into a plurality of pixels, and determining a block with less distortion. In a picture transmission apparatus that includes a motion compensator for generating a signal, quantizes the difference between the current block signal and the predicted signal, and quantizes and encodes the motion vector together with the motion vector for transmission. In addition to having a memory for outputting as a motion vector, the motion compensator is provided with a bit switch for setting a predetermined search area determined based on the previous motion vector, and the motion compensator is set to the previous time based on the bit switch. An image transmission device having a configuration for searching a frame signal. 3. A motion vector is predicted by searching a previous frame signal stored in a frame memory based on a current block signal in which an image signal is grouped into blocks for each of a plurality of pixels and determining a block with less distortion. And a motion compensator for generating a signal, wherein the difference between the current block signal and the prediction signal is obtained, then quantized, and coded together with the motion vector to be transmitted. An allowable range is set, and the motion compensator is configured to terminate the search when the distortion when searching for the previous frame signal based on the current block signal falls within the allowable distortion range. Image transmission device. 4. A motion vector is predicted by searching a previous frame signal stored in a frame memory on the basis of a current block signal in which an image signal is grouped into a plurality of pixels, and determining a block with less distortion. An image transmission apparatus that includes a motion compensator that generates a signal, quantizes the difference between the current block signal and the predicted signal, and then encodes the motion vector together with the motion vector for transmission and monitors the status of the transmission path. With a transmission path condition monitor, the distortion compensator is set to the distortion allowable range determined based on the output of the transmission path condition monitor, and the motion compensator is set to the previous frame based on the current block signal. An image transmitting apparatus, characterized in that the search is discontinued when the distortion when the signal is searched falls within the distortion allowable range.