KR100892471B1 - Motion detection device - Google Patents

Abstract

The integer pixel precision motion detection reference image data and the encoding target macroblock image data are transferred from the SDRAM 41 to the local memory 31. Integer pixel precision motion detection is performed by the integer pixel precision motion detector 21, and based on the result, the transfer area of the reference image data for 1/4 pixel precision motion detection is determined. After the transfer of the reference image data for 1/2 pixel precision motion detection, the 1/2 pixel precision motion detection by the 1/2 pixel precision motion detector 22 and the transfer of the reference image data for 1/4 pixel precision motion detection are performed. Run concurrently. The 1/4 pixel precision motion detector 23 carries out 1/4 pixel precision motion detection. As a result, the number of pipeline stages and the number of pipeline buffers can be reduced, and the pipeline processing can be speeded up.

Description

Motion Detection Device {MOTION DETECTION DEVICE}

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to moving picture coding technology, and more particularly, to a motion detection device for detecting a motion vector of a skin coded picture from a skin coded picture and a reference picture.

Today, moving picture transmission technology and accumulation technology are very important in enriching our lives.

For example, video telephony in remote locations using portable information terminals has become possible. In this video telephone, moving pictures can be transmitted to each other in synchronization with voice, so that communication with more expressive power can be realized. The transmission path of the video telephone is wireless, and the current transmission speed is 64 kbps (bit per second). In the future, it may be speeded up to about 2Mbps. However, in order to improve the image quality of the image to be transmitted at a relatively low transmission rate, a moving picture transmission technique, particularly a moving picture compression coding technique, is important.

Another important technology, the moving picture storage technology is developing year by year. In recent years, it has become possible to record a television program digitally using a DVD (Digital Versatile Disk) recorder. The number of DVD recorders is increasing year by year, and it is also a matter of time before all VHS recorders are replaced with DVD recorders. Like VHS recorders, it is an important selling point for DVD recorders to be able to record in high quality for a long time. Although the recording density of recording media (DVD-RAM, DVD-RW, Blu-ray Disc, etc.) used in a DVD recorder is improving year by year, it is not possible to record a high-vision program for a long time in high quality in the present state. In order to record a long time video in a limited area of a recording medium while maintaining the picture quality, a moving picture coding technique of encoding a video at a low bit rate without degrading the picture quality becomes important.

As a moving picture coding technique, various methods have been proposed. As standard standards of image compression technology, H. 261, H. 263 of ITU-T (International Telecommunication Union Telecommunication Standardization Division), MPEG-1, MPEG-2, MPEG-4 of ISO (International Organization for Standardization), etc. have. (MPEG stands for Moving Picture Experts Group.)

In these moving picture coding processes, the input image to be encoded includes a luminance component of 16 pixels long x 16 pixels wide, a color difference component Cb of 8 pixels wide x 8 pixels wide, and a color difference component of 8 pixels wide x 8 pixels. It is divided into macro blocks consisting of (Cr). For each macro block, the most similar block is searched from the reference picture (so-called motion detection processing is performed), and the difference with the block of the searched reference picture is obtained. Then, the difference is converted into a frequency domain and then variable length coded into a bit stream.

Among these encoding processes, a process that greatly influences image quality is a motion detection process. In the MPEG encoding processing apparatus, a motion detection unit that is an important component will first be described.

Various methods exist for motion detection, but the most typical method is block matching. The block matching method performs a pixel level operation between a macroblock of a current picture and a block of the same size as a macroblock, which is generated from a specific range in which a reference picture is located (hereinafter referred to as a search range). An evaluation value is obtained, and the position on the reference image for which the evaluation value is the best result is detected as a motion vector. As evaluation values, in general, the absolute difference value sum (SAD) or the differential square sum (SSD) is used, and the smaller the value, the higher the correlation is considered.

Conventionally, there is an example of performing motion detection hierarchically. For example, Patent Document 1 (Japanese Laid-Open Patent Publication No. 2002-218474), as a first step, performs motion vector detection with integer pixel precision in a relatively large search range in order to perform motion vector detection with 1/2 pixel precision. As a second step, a technique for detecting a motion vector with 1/2 pixel precision is disclosed around the motion vector detected in the first step and in a search range smaller than the first step.

With reference to Figs. 22 and 23, a method of motion detection according to the prior art will be described in detail.

22 is a block diagram of a conventional general motion detector. The conventional general motion detector shown in Fig. 22 is an integer pixel precision motion detector 1, 1/2 pixel precision motion detector 2, a motion compensator 3, a first local memory 4, a second local memory ( 5), third local memory 6, DMA controller 7 and SDRAM 8;

23 is a flowchart illustrating a conventional general motion detector.

In step S1 of FIG. 23, the macroblock (hereinafter referred to as the current macroblock) to be encoded is transferred from the input image stored in the SDRAM 8 to the first local memory 4.

In step S2, the motion detection range determined from the current macroblock, that is, the image data of the search range, for example, the image data of the search range of -32≤X≤ + 32, -32≤Y≤ + 32, is referred to as the SDRAM. Is transferred from (8) to the first local memory (4).

In step S3, the integer pixel precision motion detection unit 1 performs integer pixel precision motion detection for the current macroblock and the search range of the reference image transferred to the first local memory 4. In integer pixel precision motion detection, the integer pixel precision motion detection unit 1 detects a block of the same size that has the most correlation with the current macroblock from the search range by using only integer pixels, and obtains a motion vector. The motion vector is relative to the coordinates on the left end of the current macroblock and is expressed by the relative position of the coordinates on the left end of the detected block. The strength of the correlation is evaluated, for example, as the absolute difference value sum SAD or the differential square sum SSD of the luminance components in the corresponding pixels in the two blocks.

In case of performing motion detection by stratifying, in general, integer pixel precision motion detection has a larger search range than motion detection of subsequent layers. Therefore, the required memory capacity becomes large.

In order to prevent an increase in memory capacity, for example, there is a method of extracting and transferring pixels to a memory and lowering the accuracy of motion detection. Fig. 24 shows integer pixels extracted for each pixel. That is, in the example shown in FIG. 24, the pixel P2 is extracted for every pixel in the horizontal direction, and only the pixel P1 is used as a reference image. When extracted in this manner, the detection accuracy in the horizontal direction is lowered to 1/2 compared with the case where it is not extracted. However, the area for the reference picture to be secured in the first local memory 4 can be reduced to 1/2. By this method, the same search range can be searched with a small memory capacity. Alternatively, a wide range of motion detection can be performed with the same memory capacity. Which extraction method is employed is determined by a trade-off relationship between image quality deterioration due to a decrease in detection accuracy and image quality improvement due to a wider search range.

Returning to Fig. 23, in step S4, based on the motion vector MV-INT determined by the integer pixel precision motion detection in step S3, the reference image required for 1/2 pixel precision motion detection is moved from the SDRAM 8 to the second local area. Transferred to the memory 5.

As described above, when pixels of the reference image for integer pixel precision motion detection are extracted, the reference image for 1/2 pixel precision motion detection must be acquired again from the SDRAM 8. This is because, as described later, in order to calculate 1/2 pixels, adjacent integer pixels are necessarily required. In the case of performing 1/2 pixel precision motion detection with respect to 1/2 pixel of 8 points of the constant pixel precision motion vector MV-INT, the motion vector MV-INT is referred to from the reference image stored in the SDRAM 8. 18 pixels in the horizontal direction and 18 lines in the vertical direction are acquired from the coordinate position shifted by "-1" in the X direction and "-1" in the Y direction, and are transferred to the second local memory 5. do. When access to the SDRAM 8 can only be performed in units of 32 bits, pixel data that is not necessary as a reference image may be read, and image data for 24 pixels in width and 18 lines in length is read at the maximum. In some cases.

In step S5, the 1/2 pixel precision motion detection unit 2 performs 1/2 pixel precision motion detection. For example, at eight peripheral points of the motion vector MV-INT, 1/2 pixels are generated using the reference image transferred to the second local memory 5 in step S4, and the eight 1/2 pixels and The difference absolute value sum operation is performed on the integer pixel at the search center position and the current macroblock.

25 shows a 1/2 pixel generated in the vicinity of the integer pixel B. FIG. That is, 1/2 pixels a to h are generated around the integer pixel B which is the search center position. 1/2 pixel is computed as follows using integer pixels A-D, for example in the case of the simple profile of MPEG-4.

1/2 pixel f and 1/2 pixel d of FIG. 25 are computed by Formula (1) and Formula (2), respectively.

Here, R is called a rounding control, and "0" or "1" is substituted.

The 1/2 pixel precision motion detector 2 has the smallest value of the sum of the difference absolute values for a total of 9 pixels of the integer pixel B at the search center position and the eight 1/2 pixels a to h around the center. Obtain The motion vector MV-HALF with 1/2 pixel precision is calculated by adding the offset coordinates from the search center position with respect to the point where the obtained difference absolute value sum becomes the smallest to the motion vector MV-INT.

In addition, in order to improve detection accuracy, 1 / 4-pixel precision motion detection may be performed based on the motion vectors MV-HALF determined by 1 / 2-pixel precision motion detection. For example, as in the case of 1/2 pixel precision motion detection, at the peripheral eight points of the motion vector MV-HALF, a 1/4 pixel is generated using a reference image, and is generated at 1/2 pixels of the search center position and the surroundings thereof. For a total of nine pixels of one eight 1/4 pixels, the point at which the value of the sum of difference absolute values is the smallest is searched for. By adding the offset coordinates from the search center position with respect to the searched point to the motion vector MV-HALF, a motion vector with 1/4 pixel precision is calculated. In addition, in FIG.22 and FIG.23, the component and the process step for 1/4 pixel precision motion detection are abbreviate | omitted and not shown.

In step S6, for motion compensation following the motion detection, the reference image at the position indicated by the motion vector finally determined in the 1/2 pixel precision motion detection in step S5 is transferred from the SDRAM 8 to the third local memory 6. Is sent to.

In general, motion vector detection is performed on the luminance component of the pixel data. Therefore, as for the luminance component, in many cases, the reference area acquired in the second local memory 5 by 1/2 pixel precision motion detection includes an area necessary for motion compensation. In order to reduce the amount of data transmission, there is a case where the data of the second local memory 5 is transferred to the third local memory 6 and the case where the motion compensator 3 directly accesses the second local memory 5. have. However, since the color difference component is not transferred to the second local memory 5, it must be transferred from the SDRAM 8 to the third local memory 6.

In step S7, the motion compensator 3 performs motion compensation. The image data of the chrominance component obtained by the motion compensation is determined by the chrominance motion vector determined based on the motion vector of the luminance component. In the case of MPEG-4, one-half times the motion vector of the luminance component is defined as the motion vector of the color difference component. For example, the XY coordinates (0.5, 1.5) of the motion vector of the luminance component are (0.25, 0.75) when doubling, but this is rounded up to (0.5, 0.5).

As described above, the moving picture coding process is composed of a plurality of processes such as motion detection, motion compensation, DCT, variable length coding, and the like. When these processes are executed in macroblock units using one hardware resource (for example, a processor), the next macroblock process cannot be started until the process of one macroblock is finished. In such a sequential process, when the screen size or input frame rate is large, macroblock processing may not match, resulting in dropping frames.

In order to solve this problem, there is a method of providing hardware resources for each processing unit and executing macroblock processing in a pipelined manner.

26 is a flowchart of a video encoding process. As shown in Fig. 26, the general moving picture coding process includes motion detection in step S11, motion compensation in step S12, DCT / quantization processing in step S13, and variable length encoding processing in step S14. When these processes are divided into four stage pipelines, the processes as shown in FIG. 27 are obtained.

27 shows pipeline processing of video encoding. In this figure, the horizontal axis represents time, and the numerals in parentheses of the respective processes indicate macroblock numbers being processed. As shown in Fig. 27, in the pipeline processing, when the motion detection processing of the macroblock number "O" ends, the motion compensation process of the macroblock number "O" starts, and at the same time, the macroblock number "1" Processing of motion detection is started.

When the processing time of the processing with the longest processing time among the four processes shown in FIG. 27 is time T, in a pipeline processing, the macroblock streamed by the time T interval will be output. If the total time of four processes is time U, the processing time per macro block becomes time U in sequential processing, time T in pipeline processing, and U> T is self-explanatory. The throughput of block processing is improved.

However, in order to execute such pipeline processes, a pipeline buffer is required between each process. Pipeline buffers are intermediate buffers for holding data between pipelines. Therefore, pipelined work should be carried out with trade-offs in performance and cost.

28 is a flowchart of motion detection. Fig. 28 shows a process flow of motion detection of any layer in motion detection in a plurality of layers.

In step S21, motion detection of the (m-1) layer (m is a natural number of two or more) is performed. When motion detection of the (m-1) th layer performed in step S21 is performed on the extracted reference image as shown in FIG. 24, the motion vector detected in the (m-1) th layer in step S22. On the basis of this, reference picture data for motion detection of the next (m) layer should be transmitted. In step S23, motion detection of the (m) layer is performed using the transferred reference image data.

FIG. 29 shows the configuration of the pipeline for motion detection and corresponds to the motion detection in FIG. In motion detection, when searching a wide range, time is required for data transmission for acquiring data of the search range. Therefore, in the structural example of the pipeline shown in FIG. 29, the pipeline stage for data transmission is provided in stage k + 1, and productivity is improved.

As described above, while the technique according to the prior art improves the productivity of moving image processing by pipeline processing, when the hierarchy of motion detection increases, the number of pipeline stages increases, the waiting time increases, and the required pipe The drawback is that the number of line buffers increases.

(Patent Document 1)

Japanese Laid-Open Patent Publication No. 2002-218474 (Fig. 3).

(Patent Document 2)

Japanese Unexamined Patent Publication No. 2001-15872

(Tasks to be solved by the invention)

Therefore, an object of the present invention is to provide a motion detection apparatus for moving picture encoding, which can reduce the time delay in pipeline processing, suppress the occurrence of frame delay, and reduce the number of pipeline buffers. .

(Means to solve the task)

The motion detection device according to the first invention is a motion detection device that detects a motion vector hierarchically by a correlation between a reference picture and a skin coded picture, and includes a processor and a first reference picture for motion vector detection in a first step. The first motion detection means for detecting the motion vector of the first step by using the first storage means for storing, the first reference image stored in the first storage means, and the first step of detection by the first motion detection means. By using the second storage means for storing the second reference image for detecting the motion vector of the second step performed by using the motion vector and the second reference image stored in the second storage means, the motion vector of the second step is obtained. A third step for detecting the motion vector of the third step which is performed by using the second motion detecting means for detecting and the motion vector of the second step detected by the second motion detecting means. A third storage means for storing a set of images, a third motion detection means for detecting a motion vector in a third step by using a third reference image stored in the third storage means, a reference image and a skin coded image for storing And main data transfer means for controlling data transfer between the main memory means and the first memory means, data transfer between the main memory means and the second memory means, and data transfer between the main memory means and the third memory means. If reference to the motion vector of the first stage is required, the processor is further configured to transfer the data from the main storage means to the third storage means based on the motion vector detected in the first stage before the detection of the motion vector of the second stage is completed. 3 When data of the reference picture is transmitted and reference of the motion vector of the first step is unnecessary, the processor is further configured to detect the motion vector of the first step. Before, the data of the third reference image is transferred from the main storage means to the third storage means.

According to this configuration, when referring to the motion vector detected in the first step, the transmission of the reference image for motion vector detection in the third step and the execution of motion vector detection in the second step are performed simultaneously. Motion vector detection of steps can be initiated without delay. In addition, when the motion vector detected in the first step is not referenced, detection of the motion vector in the third step can be started without delay.

The motion detection device according to the second invention is a motion detection device that detects a motion vector hierarchically by a correlation between a reference picture and a skin coded picture, and includes a processor and a first reference picture for motion vector detection in a first step. The first motion detection means for detecting the motion vector of the first step by using the first storage means for storing, the first reference image stored in the first storage means, and the first step of detection by the first motion detection means. By using the second storage means for storing the second reference image for detecting the motion vector of the second step performed by using the motion vector and the second reference image stored in the second storage means, the motion vector of the second step is obtained. A second reference image for motion compensation performed using the second motion detecting means for detecting and the motion vector of the second step detected by the second motion detecting means Is a third memory means, a motion compensation means for performing motion compensation using a third reference image stored in the third memory means, main memory means for storing the reference image and the skin coded image, A data transfer control means for controlling data transfer between the first storage means, data transfer between the main memory means and the second memory means, and data transfer between the main memory means and the third memory means, wherein a reference to the motion vector of the first stage is provided. If necessary, the processor transmits the data of the third reference picture from the main storage means to the third storage means based on the motion vector detected in the first step before the detection of the motion vector of the second step is completed. In the case of dereferencing the motion vector detected in the first step, the processor starts from the main storage means before the detection of the motion vector of the first step is completed. A third storage means see 3 transmits the data of the image.

According to this configuration, when referring to the motion vector detected in the first step, since the transmission of the reference image for motion compensation and the execution of the motion vector detection in the second step are performed simultaneously, the motion compensation can be started without delay. Can be. In addition, when the motion vector detected in the first step is not referenced, the motion compensation of the third step can be started without delay.

A motion detection device according to a third aspect of the invention is a motion detection device that detects a motion vector hierarchically by a correlation between a reference picture and a skin coded picture, and includes a processor and a first reference picture for motion vector detection in a first step. The first motion detection means for detecting the motion vector of the first step by using the first storage means for storing, the first reference image stored in the first storage means, and the first step of detection by the first motion detection means. By using the second storage means for storing the second reference image for detecting the motion vector of the second step performed by using the motion vector and the second reference image stored in the second storage means, the motion vector of the second step is obtained. Second motion detecting means for detecting, main memory means for storing the reference image and skin coded image, data transfer and main memory means between the main memory means and the first memory means; And a data transfer control means for controlling data transfer between the two storage means, and the processor transfers data of the second reference image from the main storage means to the second storage means before the motion vector detection of the first step is completed. .

According to this configuration, since the transmission of the reference image for motion vector detection in the second step and the execution of the motion vector detection in the first step are performed simultaneously, motion vector detection in the second step can be started without delay.

A motion detection device according to a fourth invention is a motion detection device for detecting a motion vector by correlation between a reference picture and a skin coded picture, the motion detection device comprising: a processor and a first reference picture for storing a first step for motion vector detection in a first step; The first motion detecting means for detecting the motion vector of the first step using the first memory means, the first reference image stored in the first memory means, and the motion vector of the first step detected by the first motion detecting means. Second storage means for storing a second reference image for motion compensation to be carried out by using, motion compensation means for performing motion compensation using a second reference image stored in the second storage means, a reference image and skin encoding To control main memory means for storing an image, data transfer between the main memory means and the first memory means, and data transfer between the main memory means and the second memory means; Data transfer control means, the processor transfers data of the second reference image from the main storage means to the second storage means before the detection of the motion vector of the first step is completed.

According to this configuration, since the transmission of the reference image for motion compensation and the execution of the motion vector detection in the first stage are performed at the same time, motion compensation can be started without delay.

In the motion detection apparatus according to the fifth invention, the first motion detection means detects a motion vector of integer pixel precision.

In the motion detection apparatus according to the sixth invention, the second motion detection means detects a motion vector with 1/2 pixel precision.

In the motion detection apparatus according to the seventh invention, the third motion detection means detects a motion vector with 1/4 pixel precision.

According to these structures, it is possible to carry out stepwise from motion vector detection with integer pixel precision to motion vector detection with 1/4 pixel precision. Further, a motion detection device for performing motion vector detection with integer pixel precision, a motion detection device for performing motion vector detection with 1/2 pixel precision, or a motion detection device for motion vector detection with 1/4 pixel precision up to It can be arbitrarily configured according to the application purpose.

In the motion detection apparatus according to the eighth invention, the motion compensation means performs motion compensation of the luminance image.

According to this configuration, a motion detection device that performs motion compensation on luminance data can be realized.

In the motion detection apparatus according to the ninth invention, the motion compensation means performs motion compensation of the color difference image.

According to this configuration, a motion detection device that performs motion compensation on color difference data can be realized.

In the motion detection apparatus according to the tenth invention, the first storage means and the second storage means are mounted in a memory, and the memory size of the first storage means is larger than the memory size of the second storage means.

According to this configuration, the first motion detecting means using the first memory means can search the motion vector over a wider range than the second motion detecting means using the second memory means.

In the motion detection apparatus according to the eleventh invention, the second storage means and the third storage means are mounted in a memory, and the memory size of the second storage means is larger than the memory size of the third storage means.

According to this configuration, the second motion detecting means using the second memory means can search the motion vector over a wider range than the third motion detecting means using the third memory means.

In the motion detection apparatus according to the twelfth invention, either the data transfer control means or the second motion detection means accesses the second storage means.

In the motion detection apparatus according to the thirteenth invention, either the data transfer control means or the third motion detection means accesses the third storage means.

In the motion detection apparatus according to the fourteenth aspect, either the data transmission control means or the motion compensation means accesses the third storage means.

According to these configurations, data transfer and motion detection can be performed without providing a pipeline buffer.

In the motion detection apparatus according to the fifteenth invention, the data of the reference image in the required range is transmitted from the second storage means to the third storage means on the basis of the motion vector detected by the first motion detection means.

According to this configuration, data transfer from the main storage means to the third storage means can be omitted.

In the motion detection apparatus according to the sixteenth aspect of the invention, data of a reference image in a necessary range is transmitted from the first storage means to the second storage means on the basis of the motion vector detected by the first motion detection means.

According to this configuration, data transfer from the main storage means to the second storage means can be omitted.

(Effects of the Invention)

According to the present invention, it is possible to provide a motion detection apparatus for moving picture coding that can reduce the time delay in pipeline processing to suppress the occurrence of frame delay and also reduce the number of pipeline buffers.

1 is a block diagram of a motion detection apparatus according to the first embodiment of the present invention;

2 is a flowchart of a motion detection apparatus according to the first embodiment of the present invention;

3 is an integer pixel arrangement diagram extracted in one quarter of a reference image in Embodiment 1 of the present invention;

4 is a 1/2 pixel arrangement diagram extracted in one-fourth of a reference image in Embodiment 1 of the present invention;

5 is a 1/4 pixel layout view of a reference image in Embodiment 1 of the present invention;

6 is an explanatory diagram showing a transmission range of a reference image in Example 1 of the present invention;

7 is a configuration diagram of a pipeline of the motion detection apparatus according to the first embodiment of the present invention;

8 is a block diagram of a motion detection apparatus in accordance with a second embodiment of the present invention;

Fig. 9 is a flowchart of a motion detection device in Embodiment 2 of the present invention;

10 is a correspondence table of luminance coordinates and color difference coordinates in Example 2 of the present invention;

11 is an explanatory diagram of a transmission range of color difference data according to a second embodiment of the present invention;

12 is a pipeline configuration diagram of a motion detection apparatus according to the prior art;

13 is a configuration diagram of a pipeline of a motion detection apparatus according to a second embodiment of the present invention;

Fig. 14 is a block diagram of a motion detection device in accordance with the third embodiment of the present invention;

15 is a flowchart of a motion detection apparatus in accordance with a third embodiment of the present invention;

16 is a configuration diagram of a pipeline of the motion detection apparatus in the third embodiment of the present invention;

Fig. 17 is a flowchart of the motion detection apparatus in the fourth embodiment of the present invention;

18 is a configuration diagram of a pipeline of a motion detection apparatus in Embodiment 4 of the present invention;

19 is a block diagram of a motion detection device in Embodiment 5 of the present invention;

20 is a flowchart of a motion detection device in Example 5 of the present invention;

21 is a configuration diagram of a pipeline of a motion detection device in Embodiment 5 of the present invention;

22 is a block diagram of a conventional general motion detector;

23 is a flowchart of a conventional general motion detector;

24 is an exemplary diagram of integer pixels extracted for each pixel;

25 is an exemplary diagram of 1/2 pixels generated in the vicinity of the integer pixel B;

26 is a flowchart of a video encoding process;

27 is an illustration of pipeline processing of video encoding;

28 is a flowchart of motion detection;

29 is a configuration diagram of a pipeline of motion detection.

Explanation of symbols for the main parts of the drawings

1: integer pixel precision motion detector

2: 1 / 2-pixel precision motion detector

3: motion compensation unit 4: first local memory

5: second local memory 6: third local memory

7: DMA controller 8: SDRAM

20 processor 21 integer pixel precision motion detector

22: 1 pixel precision motion detector

23: 1 / 4-pixel precision motion detector

24: motion compensator 31, 32, 33: local memory

41: SDRAM 42: DMA Controller

Next, an embodiment of the present invention will be described with reference to the drawings.

(Example 1)

1 is a block diagram of a motion detection apparatus according to the first embodiment of the present invention. As shown in FIG. 1, the motion detection apparatus of this embodiment includes an integer pixel precision motion detector 21, a 1/2 pixel precision motion detector 22, a 1/4 pixel precision motion detector 23, and a local memory 31. 32 and 33, SDRAM 41, DMA controller 42 and processor 20.

The integer pixel precision motion detector 21 corresponds to the first motion detection means, the 1/2 pixel precision motion detector 22 corresponds to the second motion detection means, and the 1/4 pixel precision motion detector 23 is It corresponds to three motion detection means.

The local memory 31 corresponds to the first storage means, and stores the reference image data used by the integer pixel precision motion detector 21 and the image data of the encoding target macroblock. The local memory 32 corresponds to the second storage means, and stores the reference image data used by the 1/2 pixel precision motion detector 22 and the image data of the encoding target macroblock. The local memory 33 corresponds to the third storage means, and stores the reference image data used by the 1/4 pixel precision motion detector 23 and the image data of the encoding target macroblock. The SDRAM 41 corresponds to the main memory device and stores image data of the current frame and the reference frame.

The DMA controller 42 corresponds to a data transfer control means and controls data transfer between the SDRAM 41 and the local memories 31, 32, 33. The processor 20 controls the processing of the entire motion detection apparatus. 1, the solid line represents a data line, and the dotted line represents a control line.

2 is a flowchart of a motion detection apparatus according to the first embodiment of the present invention. With reference to FIG. 1, the operation | movement of the motion detection apparatus of this embodiment is demonstrated according to FIG.

In step S31, the reference image data used for integer pixel precision motion detection and the image data of the encoding target macroblock are transferred from the SDRAM 41 to the local memory 31 under the control of the DMA controller 42.

In step S32, the integer pixel precision motion detector 21 performs integer pixel precision motion detection using the reference image data transferred to the local memory 31 and the image data of the encoding target macroblock. Integer pixel motion detection is performed according to the block matching method.

In the following description, an example in which integer pixel precision motion detection of the present embodiment is performed on a reference image extracted in the horizontal direction in 1/4 is described.

FIG. 3 is an integer pixel arrangement diagram extracted as one-fourth of the reference image in Embodiment 1 of the present invention. FIG. In this figure, the pixel Fp1 of a white circle represents the integer pixel which was not extracted, and the pixel Fp2 of the black circle represents the extracted integer pixel. In this example, the reference image is extracted in a quarter in the horizontal direction. Since valid data exists in units of 4 pixels in the horizontal direction, the accuracy of motion detection in the horizontal direction is 1/4.

Numerous methods have been proposed for integer pixel precision motion detection, and typical examples thereof include full search, gradient method, diamond search, and One-at-a-Time. In the present invention, any method may be used. In addition, as the evaluation function of integer pixel precision motion detection, the sum of absolute difference values, the difference square sum, and the like of the prior art can be used.

Returning to Fig. 2, in step S33, the reference image data used for 1/2 pixel precision motion detection and the image data of the encoding target macroblock are sent from the SDRAM 41 to the local memory (by the instruction of the processor 20). 32).

In step S34, the 1/2 pixel precision motion detector 22 performs 1/2 pixel precision motion detection in the vicinity of the motion vector detected by integer pixel precision motion detection. In the 1/2 pixel precision motion detection of this embodiment, 1/2 pixel precision motion detection is performed with respect to eight 1/2 pixels around the motion vector detected by integer pixel precision motion detection.

In the following description, an example of performing the 1/2 pixel precision motion detection of the present embodiment with respect to the reference image extracted in the horizontal direction in 1/4 is described.

Fig. 4 is a 1/2 pixel arrangement diagram extracted as one-fourth of the reference image in the first embodiment of the present invention. In FIG. 4, the pixel Fp1 of a white circle represents the integer pixel which was not extracted, and the pixel Fp2 of the black circle represents the extracted integer pixel. In addition, the small white circle pixel Hp1 represents 1/2 pixel computed from the integer pixel Fp1 which was not extracted.

1/2 pixel is computed as an average of integer pixel values as mentioned above. As shown in FIG. 4, attention is paid to arbitrary search positions, and 1/2 pixels are effective in units of four pixels in the horizontal direction. In addition, even in the processing extracted in the same quarter, it can be seen that the 1/2 pixel precision motion detection requires more reference image data as compared with the integer pixel precision motion detection.

Returning to Fig. 2, in step S35, the reference image data and the image data of the encoding target macroblock used for the 1 / 4-pixel precision motion detection are sent from the SDRAM 41 to the local memory by an instruction from the processor 20. Is sent to 33.

In step S36, the 1/4 pixel precision motion detector 23 performs 1/4 pixel precision motion detection around the motion vector detected by 1/2 pixel precision motion detection.

1/4 pixel precision which is the last layer of motion detection In motion detection, in order to improve the precision of motion detection, pixel extraction is not performed.

FIG. 5 is a 1/4 pixel layout view of the reference image in Embodiment 1 of the present invention. FIG. In FIG. 5, the pixel Fp1 of a white circle represents an integer pixel, the pixel Hp1 of a small white circle represents 1/2 pixel, and the pixel Qp1 of a small black circle represents 1/4 pixel. The code addition of the pixels Fp1, the pixel Hp1, and the pixel Qp1 is typical, and not all the pixels are coded.

Calculation of 1/4 pixels is calculated | required as an average value of 1/2 pixel similarly to the case where 1/2 pixel is computed from an integer pixel. As is apparent from the arrangement of the 1/4 pixels shown in Fig. 5, the arrangement of the 1/2 pixels for calculating the 1/4 pixels, and the arrangement of the integer pixels for calculating the 1/2 pixels, the 1/4 pixel precision In motion detection, integer pixels cannot be extracted. Therefore, when the detection of 1/2 pixel precision motion from the extracted 1/2 pixel is complete | finished, the reference image data without extraction must be transmitted for 1/4 pixel precision motion detection.

By the way, as mentioned above, when waiting for the completion | finish of transfer of the reference image data for 1/4 pixel precision motion detection, and performing 1/4 pixel precision motion detection, the start of 1/4 pixel precision motion detection will be delayed, The waiting time is increased. Therefore, in step S35 shown in FIG. 2, when the integer pixel precision motion detection is completed, data for 1/4 pixel precision motion detection is transmitted so as to include a search range in 1/2 pixel precision motion detection. .

6 is an explanatory diagram showing a transmission range of a reference image in Embodiment 1 of the present invention. In FIG. 6, the code | symbol of each pixel is the same as that of FIG. 5, and abbreviate | omits description.

In the example shown in FIG. 6, it is assumed that the macroblock to be encoded is composed of 3 pixels x 3 pixels (actually, the macroblock to be encoded is composed of 16 pixels x 16 pixels). The solid edge 51 is a macroblock matched by integer pixel precision motion detection, and the position of the integer pixel precision motion vector MV-INT is given by the coordinates of the pixel Fp3 at the upper left of the edge 51. The border 52 indicated by the dotted line indicates the range of the reference image to be transmitted for 1/4 pixel precision motion detection. That is, the edge 52 is a 1/2 pixel precision motion vector from any position of the eight 1/2 pixels in the vicinity from the position of the motion vector MV-INT represented by the pixel Fp3 in the 1/2 pixel precision motion detection. Even if the detection result of MV-HALF results, the range of the pixel so that integral integer pixels necessary for generating 1/4 pixels for the next 1/4 pixel precision motion detection is necessarily included.

In this way, when the transmission range of the reference image data for 1/4 pixel precision motion detection is set to the range indicated by the edge 52, the reference image data for 1/4 pixel precision motion detection is determined for integer pixel precision motion detection. In the step where the motion vector MV-INT is determined, it can be transferred from the SDRAM 41 of FIG. 1 to the local memory 33. As a result, since the reference image data for 1/4 pixel precision motion detection can be transmitted without waiting for the result of 1/2 pixel precision motion detection, the data waiting time of 1/4 pixel precision motion detection is reduced, and the macro The latency of block processing is improved.

7 shows the configuration of a pipeline of the motion detection apparatus according to the first embodiment of the present invention. Fig. 7 shows that the pipeline of the process of the motion detection apparatus of the present embodiment is divided into a motion detection process and a reference image DMA transfer process and is configured from stage-0 to stage-4. As described above, in the motion detection apparatus of the present embodiment, in the stage-3, the reference image data for 1/4 pixel precision motion detection can be transferred simultaneously with the detection of the 1/2 pixel precision motion, The number of pipeline stages can be reduced by one step.

As described above, according to the motion detection apparatus of the present embodiment, since the number of pipeline stages can be reduced by one step, and the motion detection processing can be performed at a high speed, the frame delay can be reduced by reducing the time delay in the pipeline processing. Can be suppressed.

(Example 2)

8 is a block diagram of a motion detection apparatus in accordance with the second embodiment of the present invention. In FIG. 8, description is abbreviate | omitted by adding the same code | symbol about the component similar to FIG.

As shown in FIG. 8, the motion detection apparatus of this embodiment includes an integer pixel precision motion detector 21, a 1/2 pixel precision motion detector 22, a motion compensator 24, local memories 31, 32, 33, The SDRAM 41, the DMA controller 42 and the processor 20 are provided.

The local memory 31 corresponds to the first storage means, and stores the reference image data used by the integer pixel precision motion detector 21 and the image data of the encoding target macroblock. The local memory 32 corresponds to the second storage means, and stores the reference image data used by the 1/2 pixel precision motion detector 22 and the image data of the encoding target macroblock. The local memory 33 corresponds to the third storage means, and stores the reference image data used by the motion compensation unit 24 and the image data of the encoding target macroblock. The SDRAM 41 corresponds to the main memory device and stores image data of the current frame and the reference frame.

The DMA controller 42 corresponds to a data transfer control means and controls data transfer between the SDRAM 41 and the local memories 31, 32, 33. The processor 20 controls the processing of the entire motion detection apparatus. 8, the solid line represents a data line, and the dotted line represents a control line.

In the motion detection device of the present embodiment, motion detection is performed in two layers of integer pixel precision and 1/2 pixel precision, and motion detection of 1/4 pixel precision is not performed. In addition, the pixel of a reference image shall not be extracted by 1/2 pixel precision motion detection. After 1/2 pixel precision motion detection, motion compensation is performed.

Fig. 9 is a flowchart of a motion detection device in Embodiment 2 of the present invention.

Referring to FIG. 8, the operation of the motion detection apparatus of this embodiment will be described with reference to FIG. 9.

Transfer of reference image data for integer pixel precision motion detection and encoding target macroblock image data in step S41, integer pixel precision motion detection in step S42, and 1/2 pixel precision motion detection in step S43 The transfer of the reference image data and the encoding target macroblock image data and the 1/2 pixel precision motion detection in step S44 are performed by the steps S31 and S32 of the flowchart of the motion detection device according to the first embodiment of the present invention shown in FIG. Are the same as in step S33 and step S34, and description is omitted.

In step S44, when the 1/2 pixel precision motion detection ends, motion compensation is performed next. Motion compensation is performed on the reference picture of the luminance component and the reference picture of the chrominance component. However, in this step, reference image data of the color difference component has not yet been transferred to the local memory 33. In addition, since the reference image data area of the color difference component cannot be specified unless the motion vector of the luminance component is determined, in the prior art, the reference image data of the color difference component is transmitted after the determination of the 1/2 pixel precision motion detection. I had to.

Thus, in the motion detection apparatus of the present embodiment, in step S45, at the step where the integer pixel motion vector is determined, the transmission of the reference image data of the color difference component is started so as to include the search range of 1/2 pixel precision motion detection. . That is, as in the reference image transmission for 1/4 pixel precision motion detection in step S35 of FIG. 2 in Example 1, the reference image of the required color difference component can cope with any search result of 1/2 pixel precision motion detection. The data area is defined, and the reference image data of the color difference component of the area is transferred from the SDRAM 41 shown in FIG. 8 to the local memory 33 immediately after the determination of the motion vector in integer pixel precision motion detection.

In step S46, in accordance with the 1/2 pixel precision motion detection result in step S44, the reference image data of the luminance component and the reference image data of the chrominance component stored in the local memory 33 are read out, and motion compensation is performed. Do it.

The specific transmission method of the reference image data of the color difference component in step S45 described above will also be described.

10 is a correspondence table of luminance coordinates and color difference coordinates in Example 2 of the present invention. This correspondence table is equally applicable to the horizontal and vertical coordinates.

Since the reference image data of the chrominance component (hereinafter referred to as chrominance data) is the amount of half of the reference image data of the luminance component (hereinafter referred to as luminance data) in each of the horizontal direction and the vertical direction, One color difference data corresponds to two luminance data (one color difference data corresponds to four luminance data in the whole screen). That is, as shown in FIG. 10, the coordinate value "O" of the luminance corresponds to the coordinate value "O" of the color difference, and the coordinate values "0.5", "1", and "1.5" of the luminance correspond to the coordinate value "0.5" of the color difference. The coordinate value "2" of the luminance corresponds to the coordinate value "1" of the color difference, respectively. According to this coordinate conversion rule, for example, the XY coordinates of the color difference data corresponding to the XY coordinates (1.5, 2.5) of the luminance data are (0.5, 1.5).

In motion compensation, the color difference data to be generated corresponding to the luminance data of 16 pixels x 16 pixels of the skin-decoding macro block is 8 pixels x 8 pixels. 11 is an explanatory diagram of a transmission range of color difference data in Example 2 of the present invention. FIG. 11 shows an example of coordinate conversion from coordinates of luminance data in the horizontal direction to coordinates of color difference data for simplicity of explanation.

Now, in the luminance data, it is assumed that as a result of integer pixel precision motion detection, the position of the integer pixel precision motion vector MV-INT is obtained by the integer pixel Fp12 of the black circle. In the 1 / 2-pixel precision motion detection of the next layer, the coordinate positions where the 1 / 2-pixel precision motion vector may be detected are 1/2 pixels Hp11 and 1/2 pixels Hp12 and integer pixels Fp12 to the left and right of the integer pixel Fp12. He is himself. For example, when the X coordinate of the integer pixel Fp12 is "2", the X coordinates of the pixels Hp11, Fp12, and Hp12 for which 1/2 pixel precision motion vector may be detected are "1.5", "2", and "2.5", respectively. It becomes.

The coordinates and pixels of the color difference data corresponding to the coordinates of the luminance data are based on the coordinate conversion rule of FIG. 10, and the pixel 1/2 of the coordinate "0.5" Hp20, the integer pixel Fp21 of the coordinate "1", and the 1/2 of the coordinate "1.5". It becomes two pixel Hp21. That is, the coordinates of a pixel that may be generated as one line to eight pixels of color difference data are one of the following three cases.

In the case of including from 1/2 pixel Hp20 of coordinate "0.5" to 1/2 pixel Hp27 of coordinate "7.5", in order to generate the color difference data of (1), coordinate "8.0" from integer pixel Fp20 of coordinate "0". Integer pixel Fp28 must be transferred from the SDRAM 41 to the local memory 33.

When the integer pixel Hp28 of the coordinate "8.5" is included from the integer pixel Hp21 of the coordinate "1.5", in order to generate the color difference data of (3), the integer pixel Fp29 of the coordinate "9.0" from the integer pixel Fp21 of the coordinate "1" Up to the local memory 33 from the SDRAM 41.

From the above, in order to be able to generate all the color difference data in the cases (1), (2), and (3), the SDRAM (41) is converted from the integer pixel Fp20 at the coordinate "0" to the integer pixel Fp29 at the coordinate "9.0". To the local memory 33. By calculating in this way, it is possible to transmit the reference image data of the color difference before the 1/2 pixel precision motion detection ends.

As described above, according to the motion detection apparatus of the present embodiment, the reference image data for motion compensation can be transmitted without waiting for the result of the 1/2 pixel precision motion detection, and thus the standby for acquiring the reference image data necessary for motion compensation. The time is reduced, and the waiting time of macro block processing is improved.

Here, in order to clarify the effect of reducing the required number of pipeline stages and the number of pipeline buffers in the motion detection apparatus of the present embodiment, it will be compared with the prior art.

12 shows a pipeline configuration of a motion detection apparatus according to the prior art. At the same time, Fig. 12 shows the pipeline buffers required at each stage.

As shown in FIG. 12, in stage-0, a reference picture buffer (luminance) for holding the luminance data of the reference picture currently being transmitted is required. This is because data transfer and processing are simultaneously performed in stage-0 and stage-1 in different macroblock generations. For example, when performing integer pixel precision motion detection of the (n) -th macroblock in stage-1, reference image data for integer pixel precision motion detection of the (n + 1) -th macroblock in stage-O is performed. It is transmitting in parallel. At this time, in order not to destroy the memory area referenced by the integer pixel precision motion detection of the (n) -th macroblock, a separate buffer must be provided for data transfer in the stage-O. In addition, in order to simultaneously transfer the macroblock data (luminance data and chrominance data) of the current image used for the integer pixel precision motion detection of stage-1 in stage-O, a current macroblock buffer (luminance and color difference) is required. .

In the motion detection apparatus according to the prior art, since the transmission of the motion compensation data is executed after the 1/2 pixel precision motion detection of the stage-2 is completed, it is necessary to perform the motion compensation in the stage-3 different from the stage-2. . This is because it is functionally difficult to perform 1/2 pixel precision motion detection and motion compensation in the same stage. As a result, in the stage-2, the reference image buffer (luminance) for the luminance data transfer, the reference image buffer for the chrominance data transfer (color difference), and the reference image buffer (luminance) for the luminance data motion compensation in the stage-3 Then, a reference picture buffer (color difference) for color difference data motion compensation is required.

As described above, in the motion detection apparatus according to the prior art, four pipelines are required, and a total of 10 pipeline buffers are required.

Fig. 13 shows the configuration of a pipeline of the motion detection apparatus in the second embodiment of the present invention. According to the pipeline configuration of this embodiment, data for integer pixel precision motion detection is transmitted in stage-O, integer pixel precision motion detection is performed in stage-1, and the result is received to receive 1/2 pixel precision motion detection. Data transfer is performed. In stage-2, 1/2 pixel precision motion detection and motion compensation data (color difference data) transmission are performed in parallel, and then motion compensation is performed.

Thus, according to the motion detection apparatus of this embodiment, the transmission range of motion compensation data (color difference data) is specified by the result of integer pixel precision motion detection of stage-1, and the motion compensation data (luminance data and color difference) are specified. Since the data) can be transferred in parallel with the 1/2 pixel precision motion detection in the stage-2, the required number of pipeline stages is three stages. This is one stage less than the motion detection apparatus by the prior art shown in FIG.

13 also shows the pipeline buffers required at each stage of the pipeline. In the motion detection device of the present embodiment, the required pipeline buffer includes a reference image buffer (luminance) for luminance data of each stage, a current macroblock buffer (luminance and color difference) for luminance data and color difference data, and a color difference between stage-2. A total of seven reference picture buffers (color differences) for data. That is, in the motion detection device of the present embodiment, the number of pipeline buffers can be reduced from ten to seven of the motion detection devices according to the prior art shown in FIG.

(Example 3)

Fig. 14 is a block diagram of a motion detection apparatus in accordance with the third embodiment of the present invention. In FIG. 14, the same code | symbol is attached | subjected about the component similar to FIG. 1, and description is abbreviate | omitted.

As shown in Fig. 14, the motion detection device of this embodiment includes an integer pixel precision motion detector 21, a 1/2 pixel precision motion detector 22, local memories 31 and 32, an SDRAM 41, and a DMA controller ( 42 and a processor 20.

In the motion detection device of this embodiment, after integer pixel precision motion detection, 1/2 pixel precision motion detection is performed, and 1/4 pixel precision motion detection is not performed. In addition, the pixel of a reference image shall not be extracted by 1/2 pixel precision motion detection.

Fig. 15 is a flowchart of the motion detection apparatus in the third embodiment of the present invention.

As shown in FIG. 15, the motion detection apparatus of this embodiment transfers reference image data for integer pixel precision motion detection from the SDRAM 41 to the local memory 31 in step S51.

In step S52, integer pixel precision motion detection is performed.

In step S53, reference image data for 1/2 pixel precision motion detection is transferred from the SDRAM 41 to the local memory 32. The transfer of reference image data for 1/2 pixel precision motion detection may be performed in parallel with the transfer of reference image data for integer pixel precision motion detection in step S51, or in parallel with integer pixel precision motion detection in step S52. Also good.

The area to be transmitted of the reference image data for 1/2 pixel precision motion detection is determined without depending on the search result of integer pixel precision motion detection. The determination method is the same as the determination method of the transmission area of the reference image data for 1/4 pixel precision motion detection in Embodiment 1 of the present invention (see Fig. 6). That is, with respect to the macroblock currently being encoded, even if the integer precision motion vector comes to any position, the reference image data for 1/2 pixel precision motion detection is necessarily included so that the reference image data necessary for 1/2 pixel precision motion detection is included. Determine the area to be transferred.

In step S54, based on the search result of the integer pixel precision motion detection in step S52, 1/2 pixel precision motion detection is performed using the reference image data for 1/2 pixel precision motion detection transferred in step S53. .

As described above, according to the motion detection device of the present embodiment, since reference image data for 1/2 pixel precision motion detection can be transmitted without waiting for the result of integer pixel precision motion detection, The waiting time of the reference picture data is reduced, and the waiting time of the macro block processing is improved.

Fig. 16 shows the configuration of a pipeline of the motion detection apparatus in the third embodiment of the present invention. According to the motion detection apparatus of the present embodiment, since the reference image data transfer for 1/2 pixel precision motion detection can be performed in the stage-1, the number of pipeline stages is sufficient.

(Example 4)

The motion detection apparatus of Embodiment 4 of the present invention has the same block configuration as the motion detection apparatus of Embodiment 1 of the present invention shown in FIG. Therefore, the motion detection apparatus of this embodiment will be described with reference to FIG.

The motion detection device of this embodiment is a combination of Embodiment 1 and Embodiment 3 of the present invention, and performs integer pixel precision motion detection, 1/2 pixel precision motion detection, and 1/4 pixel precision motion detection. In addition, the motion detection apparatus of the present embodiment can perform the reference image transmission for 1/2 pixel precision motion detection without waiting for the result of integer pixel precision motion detection, and also the reference for 1/4 pixel precision motion detection. Image transmission can be started immediately after the motion vector is determined in integer pixel precision motion detection.

Fig. 17 is a flowchart of a motion detection device in Embodiment 4 of the present invention. Referring to FIG. 1, the operation of the motion detection apparatus of this embodiment will be described with reference to FIG. 17.

In step S61, the integer pixel precision motion detection reference image data is transferred.

In step S62, integer pixel precision motion detection is performed.

Simultaneously with step S62, in step S63, reference image data for 1/2 pixel precision motion detection is transferred.

In step S64, based on the search result of integer pixel precision motion detection in step S62, 1/2 pixel precision motion detection is performed using the reference image data for 1/2 pixel precision motion detection transferred in step S63. .

Simultaneously with step S64, in step S65, reference image data for 1/4 pixel precision motion detection is transferred to the data transfer area determined on the basis of the search result of the integer pixel precision motion detection in step S62.

In step S66, based on the search result of 1/2 pixel precision motion detection in step S64, 1/4 pixel precision motion detection is performed using the reference image data for 1/4 pixel precision motion detection transferred in step S65. Is done.

As described above, since the motion detection device of the present embodiment can transmit reference image data for 1/2 pixel precision motion detection without waiting for the result of integer pixel precision motion detection, the reference image for 1/2 pixel precision motion detection The latency of data is reduced. In addition, since the reference image data for 1/4 pixel precision motion detection can be transmitted without waiting for the result of 1/2 pixel precision motion detection, the waiting time of the reference image data for 1/4 pixel precision motion detection is reduced. do. As a result, according to the motion detection apparatus of this embodiment, the waiting time for macroblock processing is greatly improved.

Fig. 18 shows a configuration of a pipeline of the motion detection apparatus in the fourth embodiment of the present invention. As shown in Fig. 18, in the motion detection device of the present embodiment, the reference image data for motion detection of 1/2 pixel precision can be transferred in stage-1, and in stage-2, 1/1 is performed. Transfer of reference image data for 4-pixel precision motion detection can be performed. As a result, in the motion detection apparatus of this embodiment, the number of pipeline stages is reduced by two stages. In addition, the motion detection apparatus of this embodiment is characterized in that the waiting time of macroblock processing is determined only by the execution time of motion vector detection, so that a delay due to data transmission does not occur.

(Example 5)

Fig. 19 is a block diagram of a motion detection device in Embodiment 5 of the present invention. In FIG. 19, the same code | symbol is attached | subjected about the component similar to FIG. 1, and description is abbreviate | omitted.

As shown in FIG. 19, the motion detection device of this embodiment includes an integer pixel precision motion detector 21, a motion compensator 24, local memories 31 and 32, an SDRAM 41, a DMA controller 42 and a processor ( 20).

In the motion detection device of this embodiment, motion compensation is performed after integer pixel precision motion detection.

20 is a flowchart of a motion detection apparatus in Embodiment 5 of the present invention.

As shown in FIG. 20, the motion detection apparatus of this embodiment transfers reference image data for integer pixel precision motion detection from the SDRAM 41 to the local memory 31 in step S71.

In step S72, integer pixel precision motion detection is performed using the reference image data transferred to the local memory 31 in step S71.

In step S73, the reference image data for motion compensation is transferred from the SDRAM 41 to the local memory 32. The transfer of the reference image data is performed in parallel with the integer pixel precision motion detection in step S72.

In step S74, motion compensation is performed based on the search result of integer pixel precision motion detection in step S72, using the reference image data for motion compensation transmitted in step S73.

As described above, according to the motion detection device of the present embodiment, since the reference image data for motion compensation can be transmitted without waiting for the result of integer pixel precision motion detection, the waiting time of the reference image data for motion compensation is reduced, The latency of macro block processing is improved.

Fig. 21 shows the configuration of a pipeline of the motion detection apparatus in the fifth embodiment of the present invention. According to the motion detection device of the present embodiment, since the reference image data for motion compensation can be transmitted in stage-1, it is sufficient that the number of pipeline stages is one.

As described above, according to the motion detection apparatus of the present invention, the transfer of the reference image data for 1/2 pixel precision motion detection and the transfer of reference image data for 1/4 pixel precision motion detection are performed by the pixels of the upper layer. Since it can be performed without waiting for the result of the precision motion detection, the delay according to the transmission of the reference image data does not occur, and the waiting time of the macroblock processing is greatly improved. In addition, according to the motion detection apparatus of the present invention, the number of pipeline stages and the number of pipeline buffers can be reduced. As a result, it is possible to realize a motion detection device of a moving picture that can be processed at high speed at a smaller size and at a lower cost.

It is an object of the present invention to realize a motion detection apparatus for moving picture encoding, which can improve the waiting time of macroblock processing accompanying transmission of reference picture data and can also reduce the required number of pipeline buffers. Various applications are possible without departing from the spirit of the application.

The motion detection apparatus according to the present invention can be used, for example, in a video encoding apparatus and its application field.

Claims (16)

delete delete delete delete delete delete delete delete delete A motion detection device for detecting a motion vector by correlation between a reference picture and a skin coded picture, Processor, First storage means for storing a first reference image for motion vector detection in a first step, First motion detection means for detecting a motion vector of the first step by using the first reference image stored in the first storage means; Second storage means for storing a second reference image composed of a luminance image and a chrominance image for motion compensation performed by using the motion vector of the first step detected by the first motion detecting means; Motion compensation means for performing motion compensation using a second reference image stored in said second storage means; Main storage means for storing the reference image and the skin encoded image; Data transfer control means for controlling data transfer between the main storage means and the first storage means and data transfer between the main storage means and the second storage means. And The processor transfers data of the second reference picture from the main storage means to the second storage means before the detection of the motion vector of the first step is completed, The first storage means and the second storage means are mounted in a memory, and the memory size of the first storage means is larger than the memory size of the second storage means. Motion detection device. A motion detection device for detecting a motion vector hierarchically by correlation of a reference picture and a skin coded picture, Processor, First storage means for storing a first reference image for motion vector detection in a first step, First motion detection means for detecting a motion vector of the first step by using the first reference image stored in the first storage means; Second storage means for storing a second reference image for motion vector detection in a second step performed by using the motion vector in the first step detected by said first motion detecting means; Second motion detection means for detecting a motion vector of the second step by using the second reference image stored in the second storage means; Third storage means for storing a third reference image for motion vector detection in a third step performed by using the motion vector of the second step detected by said second motion detection means; Third motion detection means for detecting a motion vector of a third step by using a third reference image stored in the third storage means; Main storage means for storing the reference image and the skin encoded image; Data transfer control means for controlling data transfer between the main storage means and the first storage means, data transfer between the main storage means and the second storage means, and data transfer between the main storage means and the third storage means. and, When the reference of the motion vector of the first step is required, the processor is further configured to execute the first information from the main memory means based on the motion vector detected in the first step before the detection of the motion vector of the second step is completed. Transfer the data of the third reference image to the third storage means, If the reference of the motion vector of the first step is unnecessary, the processor transmits data of the third reference picture from the main storage means to the third storage means before the detection of the motion vector of the first step is completed. , The second storage means and the third storage means are mounted in a memory, and the memory size of the second storage means is larger than the memory size of the third storage means. Motion detection device. A motion detection device for detecting a motion vector hierarchically by correlation of a reference picture and a skin coded picture, Processor, First storage means for storing a first reference image for motion vector detection in a first step, First motion detection means for detecting a motion vector of the first step by using the first reference image stored in the first storage means; Second storage means for storing a second reference image for motion vector detection in a second step performed by using the motion vector in the first step detected by said first motion detecting means; Second motion detection means for detecting a motion vector of the second step by using the second reference image stored in the second storage means; Third storage means for storing a third reference image for motion vector detection in a third step performed by using the motion vector of the second step detected by said second motion detection means; Third motion detection means for detecting a motion vector of a third step by using a third reference image stored in the third storage means; Main storage means for storing the reference image and the skin encoded image; Data transfer control means for controlling data transfer between the main storage means and the first storage means, data transfer between the main storage means and the second storage means, and data transfer between the main storage means and the third storage means. and, When the reference of the motion vector of the first step is required, the processor is further configured to execute the first information from the main memory means based on the motion vector detected in the first step before the detection of the motion vector of the second step is completed. Transfer the data of the third reference image to the third storage means, If the reference of the motion vector of the first step is unnecessary, the processor transmits data of the third reference picture from the main storage means to the third storage means before the detection of the motion vector of the first step is completed. , One of the data transfer control means and the second motion detection means accesses the second storage means. Motion detection device. A motion detection device for detecting a motion vector hierarchically by correlation of a reference picture and a skin coded picture, Processor, First storage means for storing a first reference image for motion vector detection in a first step, First motion detection means for detecting a motion vector of the first step by using the first reference image stored in the first storage means; Second storage means for storing a second reference image for motion vector detection in a second step performed by using the motion vector in the first step detected by said first motion detecting means; Second motion detection means for detecting a motion vector of the second step by using the second reference image stored in the second storage means; Third storage means for storing a third reference image for motion vector detection in a third step performed by using the motion vector of the second step detected by said second motion detection means; Third motion detection means for detecting a motion vector of a third step by using a third reference image stored in the third storage means; Main storage means for storing the reference image and the skin encoded image; Data transfer control means for controlling data transfer between the main storage means and the first storage means, data transfer between the main storage means and the second storage means, and data transfer between the main storage means and the third storage means. and, When the reference of the motion vector of the first step is required, the processor is further configured to execute the first information from the main memory means based on the motion vector detected in the first step before the detection of the motion vector of the second step is completed. Transfer the data of the third reference image to the third storage means, If the reference of the motion vector of the first step is unnecessary, the processor transmits data of the third reference picture from the main storage means to the third storage means before the detection of the motion vector of the first step is completed. , A motion detection device, to which the third storage means accesses either the data transfer control means or the third motion detection means. A motion detection device for detecting a motion vector hierarchically by correlation of a reference picture and a skin coded picture, Processor, First storage means for storing a first reference image for motion vector detection in a first step, First motion detection means for detecting a motion vector of the first step by using the first reference image stored in the first storage means; Second storage means for storing a second reference image for motion vector detection in a second step performed by using the motion vector in the first step detected by said first motion detecting means; Second motion detection means for detecting a motion vector of the second step by using the second reference image stored in the second storage means; Third storage means for storing a third reference image composed of a luminance image and a chrominance image for motion compensation performed by using the motion vector of the second step detected by the second motion detection means; Motion compensation means for performing motion compensation using the third reference image stored in the third storage means; Main storage means for storing the reference image and the skin encoded image; Data transfer control means for controlling data transfer between said main memory means and said first memory means, data transfer between said main memory means and said second memory means, and data transfer between said main memory means and said third memory means; And When the reference of the motion vector of the first step is required, the processor is further configured to execute the first information from the main memory means based on the motion vector detected in the first step before the detection of the motion vector of the second step is completed. Transfer the data of the third reference image to the third storage means, If the reference of the motion vector of the first step is unnecessary, the processor transmits data of the third reference picture from the main storage means to the third storage means before the detection of the motion vector of the first step is completed. , A motion detection device, to which the third storage means accesses either the data transfer control means or the motion compensation means. A motion detection device for detecting a motion vector hierarchically by correlation of a reference picture and a skin coded picture, Processor, First storage means for storing a first reference image for motion vector detection in a first step, First motion detection means for detecting a motion vector of the first step by using the first reference image stored in the first storage means; Second storage means for storing a second reference image for motion vector detection in a second step performed by using the motion vector in the first step detected by said first motion detecting means; Second motion detection means for detecting a motion vector of the second step by using the second reference image stored in the second storage means; Third storage means for storing a third reference image for motion vector detection in a third step performed by using the motion vector of the second step detected by said second motion detection means; Third motion detection means for detecting a motion vector of a third step by using a third reference image stored in the third storage means; Main storage means for storing the reference image and the skin encoded image; Data transfer control means for controlling data transfer between the main storage means and the first storage means, data transfer between the main storage means and the second storage means, and data transfer between the main storage means and the third storage means. and, When the reference of the motion vector of the first step is required, the processor is further configured to execute the first information from the main memory means based on the motion vector detected in the first step before the detection of the motion vector of the second step is completed. Transfer the data of the third reference image to the third storage means, If the reference of the motion vector of the first step is unnecessary, the processor transmits data of the third reference picture from the main storage means to the third storage means before the detection of the motion vector of the first step is completed. , On the basis of the motion vector detected by the first motion detecting means, data of a reference image in a required range is transmitted from the second memory means to the third memory means. Motion detection device. delete

Images