JPH1013836A - Motion vector detector - Google Patents

Abstract

PROBLEM TO BE SOLVED: To reduce a processing quantity for calculation of motion vectors without losing detection accuracy. SOLUTION: A de-framing control section 202 de-frames an input image by the set number of frames. A motion vector detection section 213 uses a processing block in a processing image obtained by de-framing to retrieve a prescribed retrieval area in a reference image obtained from a frame memory 211 so as to detect a motion vector. In this case, a retrieval size selection section 212 sets a smaller retrieval area to reduce the processing quantity when the set number of de-framing is small thereby reducing the occupancy rate of a CPU and the retrieval size selection section 212 sets a larger retrieval area to attain accurate motion detection when the set number of de-framing is large.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a motion vector detecting device suitable for use in motion compensation interframe coding, which is a compression coding method for moving picture information.

[0002]

2. Description of the Related Art In recent years, the development of image compression encoding technology and the spread of digital communication lines have been remarkable, and service regulations and protocol regulations for AV (Audio Visual) services such as TV conference systems, multimedia multiplex frame configuration regulations and the like have been made. And the TV
Various multimedia communication terminals including a telephone device and a desktop TV conference system have been proposed. In these multimedia communications, not only communication using moving images and voices, but also data communication and joint work at remote points by sharing applications are realized.

As is well known, a direct digital transmission of TV image information requires a transmission speed of several hundred Mbps. Therefore, various compression coding schemes are used to reduce the transmission speed and further reduce the transmission cost. Has been proposed. As a compression coding method for moving picture information in these moving picture communication terminals, a temporal redundancy is removed by a motion compensation inter-frame prediction coding using a correlation in a time direction, and an orthogonal transform code using a correlation in a space direction. Hybrid coding schemes that remove spatial redundancy by coding are often used.

Next, the principle of motion-compensated inter-frame coding used for compression coding of moving picture information will be briefly described. If the target object moves between frames, a structure similar to the coding processing part (macroblock) is not located at the same position of the previous coded frame (reference frame) but at a distance corresponding to the movement of the object. Will exist. Therefore, the motion vector of the object is estimated from the previous frame and the current frame, and the difference between similar blocks and the motion vector are encoded. The prediction error value in the motion-compensated inter-frame prediction coding is smaller than the prediction error value in the simple inter-frame prediction coding, so that efficient coding can be performed even for an image having a large motion.

FIG. 7 is a schematic diagram for explaining motion compensation.
A motion vector is detected by searching using a processing block indicated by a black portion in the processing frame as the current frame as a search area near the processing block in the reference frame as the previous frame. In the detection of the motion vector, an evaluation value such as a sum of absolute differences between corresponding pixels is calculated for the processing block of the current frame and the search area centered on the processing block position of the previous encoded frame (reference frame). , A vector corresponding to a block having a minimum value (maximum correlation) is generally estimated as a motion vector. As this procedure, a full search procedure for performing an evaluation value operation on all block patterns in the search area of the reference frame is often used. For example, ITU-T
Recommendation H. In H.261, a motion vector search area of a maximum of 47 pixels × 47 pixels is used for a macroblock of 16 pixels × 16 pixels, and the horizontal and vertical components of the motion vector are specified as integers within +/− 15, respectively. Have been. In MPEG and the like, motion compensation based on 1/2 pixel search accuracy is specified.

[0006]

Conventionally, moving image compression processing has generally been performed by hardware such as a DSP or a dedicated LSI. However, as the processing capacity of the CPU has been greatly improved, a form in which part or all of the moving image encoding processing is realized by software by the host CPU is increasing. However, the amount of processing related to moving image compression is extremely large, and when executing other applications in parallel, the transmission speed of the line exceeds the moving image encoding processing speed, and the frame rate of the transmitted image increases. Smooth (smooth motion) video communication may be hindered by a plateau, or the CPU may be occupied by video compression processing, which may have a significant adverse effect on the operation speed of other applications. There was a problem.

In general, a motion vector detection operation represented by a sum of absolute differences between block pixels requires a large amount of operation even when compared with other image encoding processes represented by an orthogonal transformation operation. Need. In the above-described full search procedure, when a motion vector search operation with one pixel accuracy is performed for a search area of 47 pixels × 47 pixels, 2
The sum of 56 absolute differences is executed 961 times. As a means for avoiding this, a method is used in which the number of calculations is reduced by always reducing the search area.
For example, a motion vector within a horizontal / vertical direction +/− 7 is calculated using a motion vector search area of 31 pixels × 31 pixels. However, when the search area is narrowed, there is a disadvantage that a correct motion search cannot be executed when a large motion occurs between the encoded screens.

As another avoiding means, a method of simplifying a search procedure to reduce the amount of calculation is used. In this method, a step search is performed instead of a full search procedure. However, if an erroneous vector is selected in a search step in the middle, an optimal search cannot be performed. Puts)
There was a drawback that there was fear.

SUMMARY OF THE INVENTION The present invention has been made in view of the above problems, and has as its object to provide an excellent motion vector detecting device capable of efficiently reducing the amount of vector calculation processing without impairing the accuracy of motion search. It is in.

[0010]

According to the present invention, a search for a reference screen located in the vicinity of a processing block by a predetermined number of screens from the processing screen using image data of a predetermined processing block in the processing screen. In a motion vector detecting device for detecting a motion vector by searching image data of an area, a frame drop control means for dropping a frame between the reference screen and the processing screen by a set number of frames, Search range control means for controlling the size of the search area in the reference screen in accordance with the number of screens dropped by the control means, and the search block in the search area determined by the processing block and the search range control means A motion search means for evaluating a correlation with each block pattern and outputting a motion vector is provided.

[0011]

According to the above arrangement, it is possible to control the motion vector search range according to the period between the screen and the motion search target screen.

[0012]

Embodiments of the present invention will be described below in detail with reference to the drawings. First, a first embodiment will be described. FIG. 1 is a schematic block diagram showing an example of a circuit configuration of a part of a moving picture coding apparatus according to the present invention. Reference numeral 201 denotes a frame memory # 1 unit, which is an image input device (not shown)
Frame for temporarily storing digital input video from
Memory. Reference numeral 202 denotes a frame drop control unit, which is a frame memory # 1 unit 201 in accordance with an instruction from the frame drop number determination unit 218.
Of the moving image information stored in. 203
Denotes a subtraction unit that performs a subtraction process between an input pixel value and a prediction pixel value, and selects and outputs a prediction error value and an input value that are the result of the subtraction. A DCT unit 204 performs a discrete cosine transform (hereinafter, referred to as DCT) process, which is a type of orthogonal transform, on the input prediction error data (or input data). A quantization unit 205 selects a quantization step size according to an instruction from the encoding control unit 217, and quantizes the input DCT coefficient data. Reference numeral 206 denotes a variable-length coding unit that performs variable-length coding on the quantized DCT coefficient data. Reference numeral 207 denotes a transmission buffer unit.

An inverse quantization unit 208 inversely quantizes the quantized DCT coefficient data. Reference numeral 209 denotes an inverse DCT unit that performs an inverse DCT transform process on the input DCT coefficient data. Reference numeral 210 denotes an adding unit that adds a corresponding predicted value as needed. Reference numeral 211 denotes a frame memory # 2, which is a frame memory for temporarily storing a reproduced image for motion compensation inter-frame prediction.

Reference numeral 212 denotes a motion vector search block 220
And a search size selection unit for selecting the size of the motion search area based on the output of the number of dropped frames determination unit 218. 213
Reference numeral denotes a motion vector search unit of the motion vector search block 220, which detects a motion vector for the immediately preceding reproduced image stored in the frame memory # 2 of the processing block of the current frame. Reference numeral 214 denotes a motion compensation unit, which is a motion vector detection unit 213 in the motion vector search block 220.
According to the motion vector detected in step (1), the corresponding pixel value data in the immediately preceding reproduction frame is selected and output as a prediction block. A filter unit 215 is a low-pass filter for a block for which motion compensation inter-frame prediction is performed. A delay unit 216 delays the prediction block by the processing time of the inverse DCT unit 209 from the subtraction unit 203 and adjusts the timing of the addition process in the addition unit 210.

Reference numeral 217 denotes an encoding control unit which performs various encoding controls such as quantization step size selection and significant block determination based on a control signal from the outside and the buffer accumulation amount of the transmission buffer unit 207. . Reference numeral 218 denotes a frame drop number determining unit in the encoding control unit 217, which determines the number of dropped frames (frame skip) in the same manner as in other encoding controls.

Next, the operation of the moving picture coding apparatus having the above configuration will be described. Fixed frame from outside
The moving image information input at the rate is stored in the frame memory #
It is stored in the first unit 201. The frame drop control unit 202 performs frame drop processing according to the instruction of the frame drop number determination unit 218, selects an encoded frame, and outputs it to the subtraction unit 203 and the motion vector detection unit 213 in a predetermined block unit. The subtraction unit 203 receives input pixel value data and prediction value data in block units, generates a difference value (prediction error) between the pixel value data and the prediction value data, and selects a prediction error block and an input block. And outputs it to DCT section 204. The DCT unit 204 performs DCT (Discrete Cosine Transform) processing, which is a type of orthogonal transform, on the output data of the subtraction unit 203 in block units, and outputs DCT coefficient data to the quantization unit 205. The quantization unit 205 quantizes the input DCT coefficient with the selected step size according to the quantization characteristic control from the encoding control unit 217. The variable length coding unit 206 performs variable length coding on the quantized DCT coefficients and outputs the result to the transmission buffer unit 207. The transmission buffer unit 207 is configured by a buffer memory, buffers variable-length encoded data, and sends the buffer storage amount and the encoded data amount in frame units to the encoding control unit 217.

On the other hand, the inverse quantization unit 208 includes the quantization unit 2
The quantization output unit 05 receives the quantization output, performs inverse quantization using the quantization step size selected by the quantization unit 205, and outputs DCT coefficients. In the inverse DCT unit 209,
The DCT coefficient output of the inverse quantization unit 208 is subjected to inverse DCT transform processing, and is output to the addition unit 210. In addition section 210, inverse D
The CT unit 209 adds the corresponding prediction data output from the delay unit 216 to the prediction error data subjected to the inverse DCT processing, and outputs a reproduced image. Frame memory #
The second unit 211 is a frame memory for motion-compensated inter-frame prediction composed of a frame memory for two frames, for example. Therefore, the pixel value data of the immediately preceding coded frame (reproduced image) is output according to the instruction of the motion compensation unit 214.

The search size determining section 212 selects the size of the search area in the motion vector detecting section 213 based on the number of dropped frames (frame skip) determined by the dropped frame number determining section 218 (details will be described later). Do). The motion vector detecting unit 213 reads out the pixel value data of the pre-encoded frame (reproduced image) near the processing block position of the current frame from the frame memory unit 211 as a motion vector search window (search area).
By performing a block matching operation in a search procedure according to the data storage amount of the transmission buffer unit 205, a motion vector is detected, and the detected motion vector is
Output to

In accordance with the motion vector detected by the motion vector detecting section 213, the motion compensating section 214 reads out pixel value data for each block of the corresponding pre-encoded frame from the frame memory # 2 section 211 and outputs it. The filter unit 215 is a low-pass filter for the purpose of reducing discontinuity at a block boundary due to the motion compensation. The filter unit 215 performs a filtering process on the data on which the motion compensation has been performed, and generates a subtraction unit 203 as predicted value data. And to the delay unit 216.

The encoding control unit 217 controls the timing of the entire image encoding apparatus and, at the same time, adaptively controls the quantization characteristic control signal based on the amount of data stored in the transmission buffer unit 207 and the amount of compressed data for each frame. , Adaptive coding control such as selection of a quantization step size and control of a variable threshold in the quantization unit 203. Also, by constantly monitoring the entire image encoding device or various control signals from the outside, it is possible to respond to changes in the image input device and various settings of the user. The frame drop number determination unit 218 in the encoding control unit 217 determines the frame drop number (frame skip) based on the data storage amount of the transmission buffer unit 207 and the compressed data amount for each frame in the same manner as described above. It notifies the drop control unit 202 and the search size selection unit 212.

Next, encoding control in the encoding control unit 217, the motion vector detecting unit 213, and the subtracting unit 203 will be described. Regarding the quantization characteristics in the quantization unit 203, if the quantization step size is set small,
Although the image quality is improved, significant data increases and leads to an increase in the number of transmission bits. If a large quantization step size is set, the amount of generated data decreases but the image quality deteriorates. The image quality and the ability to follow movement are contradictory, and the pursuit of high image quality inevitably deteriorates the ability to follow movement. Therefore, it is necessary for the encoding control unit 217 to constantly monitor the amount of data stored in the transmission buffer unit 207, and to set it appropriately and efficiently.

The number of dropped frames determination unit 218 is based on the notification of the buffer accumulation amount from the transmission buffer unit 207 and the amount of data generated in one screen, and is combined with an external image quality setting or the like to drop frames (frame skipping). ) Determine the number. FIG. 2 is a conceptual diagram showing the state of dropped frames.

The motion vector detecting section 213 determines whether or not to perform motion compensation, that is, whether to encode the difference from the motion-compensated block based on the evaluation result of the block matching operation (INTER + MC )
Alternatively, simply encode the difference between the reference frame and the same-position block without performing motion compensation (INTER + no
(MC mode) is determined. Although motion-compensated inter-frame difference information cannot be larger than simple inter-frame difference information, the amount of code is increased because motion compensation requires vector information. To prevent compression encoding,
The ON / OFF control of the motion compensation is executed adaptively. Specifically, it is determined according to the evaluation value at the same position as the processing block and the evaluation value at the motion compensation position (the block position having the maximum correlation).

The subtraction unit 203 compares the evaluation result of the prediction error block with the evaluation result of the input block using a predetermined evaluation function to encode the prediction error block (INTER + MC / noMC mode). It is determined whether the block is to be coded (INTRA mode), and one of the blocks is output to the subsequent stage, so that the compression coding is performed by adaptively selecting the inter-frame coding method and the intra-frame coding method. .

Next, the control for selecting the motion vector search size in the motion vector search block 220 characteristic of the present invention will be described in detail. First, the operation of detecting a vector with one-pixel accuracy in the motion vector detecting unit 213 will be briefly described. The motion vector is detected by calculating the absolute value of the difference between the corresponding pixels in the search area centered on the processing block position of the current frame and the processing block position of the previous coded frame (reference frame), and summing them (the absolute difference value). The sum is used as a correlation evaluation value, and a vector corresponding to a block whose evaluation value takes a minimum value (correlation is maximum) is estimated as a motion vector.

The size of the motion vector search area is determined based on the number of dropped frames. Hereinafter, a specific example will be described in detail. Here, the block which is the unit of the encoding process is 16
× 16 pixels and a motion compensation range (search range) are selected from three types. FIG. 3 is a diagram illustrating three types of search sizes. FIG. 3A shows the maximum search range +/- 1.
5 (the search area is a 47 × 47 pixel range). FIG.
(B) is +/− 11 (the search area is a 39 × 39 pixel range)
It is. FIG. 3C shows +/− 7 (the search area is 31 × 31).
Pixel range).

Next, the size control of the motion vector search area based on the number of dropped frames (frame skip) in the search size selection unit 212 will be described in detail. The search size selection unit 212 selects the size of the motion search area based on the number of dropped frames (frame skip) determined and notified by the dropped frame determination unit 218. Here, the number of dropped frames is the number of frames between the immediately preceding encoded (transmitted) frame and the processing frame. In other words, how many frames (time) should be searched for movement between frames,
According to such information, the search size in the processing frame is determined. Utilizing the fact that the range (size) in which the imaging target can move during that period varies depending on the size of the temporal separation from the immediately preceding encoded frame to be searched for motion (the number of dropped frames), If the number of dropped frames is small, the search area is limited to a small size.
If the number of dropped frames is large, the search area size is selected and controlled so that the search area is sufficiently large.

FIG. 4 is a flowchart showing an example of the control procedure of the motion vector search area size. Based on the buffer storage amount information from the transmission buffer unit 207, the generated compression code amount information of one encoded frame, and the user setting information from the outside, the number of dropped frames determination unit 218 in the encoding control unit 217 determines The number of dropped frames (frame skips) is determined (Sa1). The determined number of dropped frames is reported to the dropped frame controller 202 and also to the search size selector 212 (Sa2). The search size selection unit 212 selectively controls the size of the motion search area according to how many dropped frames exist between the next processing frame and the immediately preceding processing frame (dropped frame number information). I do.

That is, when the number of dropped frames S> 9 (S
In a3), 47 pixels × 47 pixels (motion vector +/− 15: FIG. 3A) are selected as a search range (Sa).
4) If the number of dropped frames 9 ≧ S> 2 (Sa5),
39 pixels x 39 pixels (motion vector + /
-11: FIG. 3 (b)) is selected (Sa6), and the number of dropped frames S
If ≦ 2, 31 pixels × 31 pixels (motion vector +/− 7: FIG. 3C) are selected as the search range (Sa7). The motion vector detection unit 213 performs a motion vector detection operation on the motion search size (motion compensation size) selected and controlled by the search size selection unit 212 (Sa8).

That is, when the temporal separation from the immediately preceding coded frame to be subjected to motion search is small (the transmission frame rate is high), the amount of image coding processing that must be executed by the CPU per unit time Is inevitably increased, so that an efficient motion search process is performed by limiting the motion search range to a small range, utilizing the narrow range in which the imaging target can move between the immediately preceding encoded frame and the current frame. By doing so, it is possible to reduce the occupation amount of the CPU capacity or further improve the transmission image frame rate. On the other hand,
When the temporal separation from the immediately preceding coded frame is large (the transmission frame rate is low), a motion search is performed on a sufficiently large search area, so that accurate motion compensation is performed and efficient compression is performed. Encoding can be realized.

As described above in detail, according to the present embodiment, prior to the execution of the motion vector search calculation, the number of frames dropped between the reference screen (the immediately preceding encoded transmission screen) and the processing screen is determined. By controlling the size of the motion search (compensation) area, the shorter the period (higher frame rate) between the reference screen and the processing screen, the shorter the distance (range) that the target object can move during that period. Using this, the search range is reduced as the number of dropped frames decreases (as the frame rate increases), thereby executing a motion vector search with a reasonable motion search range (motion compensation range).
That is, when the number of dropped frames is small, the search area is set to be relatively small so that the further improvement of the frame rate is not hindered by the limit of the processing capacity of the CPU, and the amount of motion vector search processing is reduced. A reduction in the CPU occupation rate (operating rate) due to the reduction is realized. On the other hand, when the number of dropped frames is large, the search area is set relatively large and the motion search processing is performed for a sufficiently large search range. By realizing accurate motion detection,
An efficient motion vector search can be performed.

Next, a second embodiment will be described.
In the first embodiment, the number of dropped frames for three types of search areas is based on the dropped frames representing the time lag between the immediately preceding encoded transmission frame (reference frame) and the processing frame. The process of selecting and controlling a smaller search range (size) as the size becomes smaller and executing a motion vector search in units of one pixel has been described. However, the present invention is not limited to this, and the same effect can be exerted by combining the above-described control of the search range with the selection control of the search accuracy according to the second embodiment.

FIG. 5 is a schematic block diagram showing an example of a circuit configuration of a part of the image encoding apparatus according to the second embodiment. The basic configuration is the same as that of the first embodiment in FIG.
The same blocks as in FIG. 1 are given the same numbers. Next, selection control of the motion vector search size and the search accuracy in the motion vector search block 620 characteristic of the present embodiment will be described in detail. First, the vector detection operation in the motion vector detection unit 613 will be briefly described. First, a motion search process with one-pixel accuracy is performed in the same manner as described in the first embodiment.
The absolute value of the difference between the corresponding pixels is calculated, the sum of the differences (sum of the absolute values of the differences) is used as a correlation evaluation, and a block having the minimum evaluation value (the maximum correlation) is selected (one pixel). Precision motion vector search). Then, 1/2
For the motion search with pixel accuracy, four interpolated blocks which are shifted by half a pixel positioned between the block detected with one pixel accuracy and the four blocks near the block are generated, and the detected block and the four interpolated blocks are generated. With respect to the combined five blocks, the correlation between the block and the processing block is evaluated, and the block having the highest correlation is selected (a motion vector search with 1/2 pixel precision). The vector for the block selected by the above two-stage search is estimated as a motion vector with 1/2 pixel accuracy.

The search accuracy of the motion vector and the size of the search area are determined based on the number of dropped frames. Hereinafter, a specific example will be described in detail. Here, it is assumed that a block which is an encoding processing unit is 16 × 16 pixels, two types of motion search accuracy are selected, and a motion compensation range (search range) is selected from the three types described with reference to FIG. That is, FIG. 3A shows a maximum search range of +/− 15 (the search area is a 47 × 47 pixel range). FIG. 3B shows + /
−11 (the search area is a 39 × 39 pixel range). FIG.
(C) is +/− 7 (the search area is a 31 × 31 pixel range). Regarding the motion search accuracy, one pixel accuracy search and 1
/ 2 pixel precision search is selectively used.

Next, the size control of the motion vector search area based on the number of dropped frames (frame skip) in the search size selection unit 212 will be described. As in the first embodiment, the search size selection unit 212
The size of the motion search area is selected based on the number of dropped frames (frame skip) determined and notified by the dropped frame determination unit 218. Utilizing the fact that the range (size) in which the imaging target can move during that period varies depending on the size of the temporal separation from the immediately preceding encoded frame to be searched for motion (the number of dropped frames), If the number of dropped frames is small, the search area is limited to a small size. If the number of dropped frames is large, the search area size is controlled so that the search area is sufficiently large.

Next, the selection control of the motion vector search precision based on the number of dropped frames (frame skip) in the search precision selection section 619 will be described in detail. The search accuracy selection unit 619 controls the motion search area selection (1 pixel accuracy or 1/2 pixel accuracy) based on the number of dropped frames (frame skip) determined and notified by the dropped frame number determination unit 218. Do. In order to supplement the size control of the motion search range described above, when the number of dropped frames is sufficiently small, the search accuracy is reduced to one-pixel accuracy, that is, 1/2 that is executed following the one-pixel accuracy search. Control is performed so as not to perform the pixel accuracy search process.

FIG. 6 is a flowchart showing an example of a control procedure of the motion vector search accuracy and the area size.
The encoding control unit 217 is based on the buffer accumulation amount information from the transmission buffer unit 207, the generated compression code amount information of one encoded frame, and external user setting information.
The number of dropped frames (frame skip) is determined by the dropped frame determination unit 218 (Sb1). The determined frame drop number information is notified to the frame drop control unit 202 and also to the search size selection unit 212 and the search accuracy selection unit 619 (Sb2). In the search size selection unit 212 and the search accuracy selection unit 619, the motion search area is determined according to how many dropped frames exist between the next processing frame and the immediately preceding processing frame (dropped number information). The size and the search precision of each are selectively controlled.

That is, when the number of dropped frames S> 9 (S
In b3), 47 pixels × 47 pixels (motion vector +/− 15: FIG. 3A) are selected as the search range (Sb
4), 1/2 pixel accuracy is selected as the search accuracy (Sb)
5). If the number of dropped frames is 9 ≧ S> 2 (Sb6), 39 pixels × 39 pixels (motion vector +/− 11: FIG. 3 (b)) are selected as the search range (Sb7), and the search accuracy is determined. The half-pixel accuracy is selected (Sb8). If the number of dropped frames S ≦ 2, the search range is 31 pixels × 3.
One pixel (motion vector +/− 7: FIG. 3C) is selected (Sb9). One pixel accuracy is selected as the search accuracy (Sb10). In the motion vector detection unit 513,
For the motion search size (motion compensation size) selected and controlled by the search size selection unit 212, the search accuracy selection unit 6
The motion vector detection calculation is executed with the motion search accuracy selected and controlled by S19 (Sb11).

That is, when the time difference from the immediately preceding encoded frame to be subjected to motion search is small (the transmission frame rate is high), the amount of image encoding processing that must be executed by the CPU per unit time Is inevitably increased, so that the motion search range is limited to a small range by utilizing the narrow range in which the imaging target can move between the immediately preceding encoded frame and the search accuracy is further reduced, thereby improving the efficiency. Perform a typical motion search process. As a result, the occupation amount of the CPU capacity is reduced, or the transmission image frame rate is further improved. On the other hand, when the temporal separation from the immediately preceding coded frame is large (the transmission frame rate is low), by executing a motion search with sufficient search accuracy on a sufficiently large search area, accurate motion Compensation is performed, and highly efficient compression encoding can be realized.

As described in detail above, according to the present embodiment, in the motion vector detection processing, the period between the immediately preceding encoded frame (reference frame) to be searched for motion and the processing frame is short ( The higher the frame rate, the more distance (range) the target object can move during that period
Utilizing the fact that the search range is reduced, the search range is controlled to be reduced as the number of dropped frames (frame period) is reduced (as the frame rate is increased), and the search accuracy is further reduced. Thus, when the frame rate is sufficiently high, the motion search range is set to be small, the search accuracy is further reduced, the increase in the amount of compression encoding processing is suppressed, and the increase in the frame rate is prevented from being hindered. In addition, the occupancy of the CPU is reduced by reducing the amount of motion vector search processing. On the other hand, when the frame rate is low, the CPU has sufficient processing power, and therefore, by performing a motion search process with a high search accuracy in a sufficiently large search range, correct motion detection can be realized.

In the first and second embodiments, the case where three types of search ranges are selected has been described. However, only two types of search ranges may be selected, or four or more types of search ranges may be selected. May be. Also, the size of the search range is not limited to the above embodiments.
In the second embodiment, only the case where two types of search accuracy are selected has been described, but it goes without saying that three or more types of search accuracy may be selected. Further, search accuracy other than 1-pixel accuracy and 1 / 2-pixel accuracy may be used. Also, the search method with half-pixel accuracy is not limited to the method described in the second embodiment.

In the above embodiments, the full search procedure using the sum of absolute differences between corresponding pixels as an evaluation function for correlation has been described as an example. It may be combined with selection control with a full search procedure. The present invention is also applicable to a case where the present invention is achieved by supplying a program to a system or an apparatus.

[0043]

As described above, according to the present invention,
In the motion vector detection processing, the shorter the period (the higher the frame rate) between the immediately preceding coded frame (reference frame) as the motion search target and the processing frame, the more the distance (range) in which the target object can move during that period. ) Is used to control the search range according to the number of dropped frames (frame period). If the frame rate is sufficiently high, the motion search range is set to a small value, and the compression code is set. While suppressing the increase in the amount of data processing to avoid hindering further frame rate improvements,
By reducing the amount of vector search processing, the CPU occupancy is reduced, and when the frame rate is low, the CPU has enough processing power to perform motion search processing with a sufficiently large search range, thereby enabling accurate motion detection. Can be done.

Further, by controlling the search accuracy in accordance with the set number of dropped frames, it becomes possible to further control the motion vector search accuracy in accordance with the period from the screen to which the motion is to be searched. .

[Brief description of the drawings]

FIG. 1 is a block diagram illustrating a schematic configuration of an image encoding device according to a first embodiment of the present invention.

FIG. 2 is a configuration diagram illustrating dropped frames (frame skipping).

FIG. 3 is a configuration diagram showing types of search areas according to the first and second embodiments of the present invention.

FIG. 4 is a flowchart illustrating an operation according to the first embodiment of the present invention.

FIG. 5 is a block diagram illustrating a schematic configuration of an image encoding device according to a second embodiment of the present invention.

FIG. 6 is a flowchart illustrating an operation according to the second exemplary embodiment of the present invention.

FIG. 7 is a configuration diagram illustrating motion compensation between frames.

[Explanation of symbols]

 201 frame memory 202 frame drop control unit 212 search size selection unit 213 motion vector detection unit 218 frame drop number determination unit 613 motion vector detection unit 619 search accuracy selection unit

Claims (4)

[Claims] 1. A motion by searching image data of a search area near a processing block position on a reference screen separated by a predetermined number of screens from the processing screen using image data of a predetermined processing block in the processing screen. In the motion vector detecting device for detecting a vector, the frame between the reference screen and the processing screen is frame-dropped by a set number of frames. A search range control means for controlling the size of the search area in the reference screen in response thereto; and evaluating a correlation between the processing block and each block pattern in the search area determined by the search range control means. A motion vector detecting device comprising: a motion search unit that outputs a motion vector. 2. A search precision control means for determining a motion vector search precision in accordance with the number of frames dropped by the frame drop control means, wherein the motion search means comprises a search determined by the search precision control means. The motion vector detecting device according to claim 1, wherein a motion vector detecting process is performed based on accuracy. 3. The search range control means sets and controls the size of the search area to be relatively small when the number of frames dropped by the drop control means is small. The motion vector detecting device as described in the above. 4. The motion vector according to claim 2, wherein said search range control means controls said search accuracy to be relatively low when the number of frames dropped by said drop control means is small. Detection device.