JP2002112274A - Processing time adaptive image coding method and recording medium for its program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a processing time adaptive image coding method that can encode a moving picture in real time even when the processing time differs depending on an execution environment of an encoding program. SOLUTION: A coding object time that is a time for encoding one frame is set on the basis of a predetermined frame rate prior to encoding and motion search instruction number information that is number of processing instructions for motion search executed by an arithmetic unit is stored according to a motion search method preset before the encoding. In the case of the encoding, a coding time of each frame is measured, a motion search object time for motion search of a succeeding frame is set on the basis of the preset coding object time and the measured coding time after the end of encoding of one frame, the motion search method is decided from the motion search object time according to the motion search instruction number information stored in advance and the motion search processing is controlled by the method.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a moving picture coding technique, and more particularly to a technique for real-time coding by switching search precision even when the coding processing time varies depending on the execution environment of the coding processing. The present invention relates to a processing time adaptive image encoding method capable of encoding a moving image.

[0002]

2. Description of the Related Art In moving picture coding, a motion-compensated inter-frame predictive coding method is generally used to improve coding efficiency. In this method, a difference (referred to as a prediction error) of image information between corresponding pixels between an original image of a current frame and a reference image formed from a decoded image of another frame encoded earlier is referred to. Encode. The reference image can be created by using a past or future frame or by affine-transforming a past or future frame.

In encoding prediction errors, first, the prediction errors are subjected to frequency transformation such as discrete cosine transform (DCT), and the coefficients are quantized, followed by variable length encoding. When the prediction error of the position shifted in parallel is smaller than the prediction error between pixels at the same position in the screen, the prediction error of the position shifted in parallel is encoded in order to improve encoding efficiency. The translation amount is encoded as a motion vector.

In general, a screen is divided into small areas, for example, by macroblocks of 16 pixels vertically and horizontally, and a motion vector is set for each of the small areas. For example, a pixel at a position that is similarly translated by a motion vector with respect to a pixel in a macroblock is set, and DCT is performed on a prediction error between these pixels. Also, instead of setting the same motion vector for each pixel in the small area, there is also a method of calculating the parallel movement amount of each pixel using the motion vector of the adjacent small area.

Reference: GJ Sullivan and RLBaker: "
Motion Compensation for video compression using co
ntorol grid interpolation ", IEEE ICA SSP '91, pp.27
13-2716,1991.

If the translation amount is not an integer, there is no pixel at the corresponding position in the reference image, and thus the value is calculated from the image information of the peripheral pixels of the reference image.

[0007] The process of obtaining the amount of translation is called "motion search". In general, a screen is divided by a macroblock of 16 pixels in length and width, and a motion vector is obtained for each macroblock. As the evaluation function of the prediction error for all the pixels in the macroblock, the sum or the variance of the absolute value differences of the image information of all the pixels in the macroblock is used. If a motion vector at a position where these values are small is selected, the encoding efficiency of the prediction error will be high. Therefore, a position where these values are minimized is determined.

Further, as the search range is enlarged, the number of motion vector candidates increases, so that a motion vector with high coding efficiency can be obtained. Video coding system In 263, the search range is shifted by a maximum of 16 pixels in the vertical and horizontal directions due to the restriction on the expression of the motion vector. This is shown in FIG. The absolute value difference sum and the variance are calculated while shifting the search range one pixel at a time, and the position where these values are minimum is obtained.

However, as the search range is enlarged, the amount of calculation for motion search increases. Various techniques have been used to reduce the amount of computation for motion search. The following are three examples.

[Method 1]: The number of pixels for calculating the sum of absolute differences and the variance is reduced. For example, calculation is performed with 64 pixels out of 16 pixels in the vertical and horizontal directions (total of 256 pixels) in the macro block. FIG. 2 shows an example in which 64 pixels are selected in a macro block.
Shown in In FIG. 2, when a pixel indicated by a black circle is selected, the calculation is performed for 64 pixels.

[Method 2]: The number of positions for calculating the sum of absolute differences and the variance within the search range is reduced. For example, the search is performed while shifting by two pixels. FIG. 3 shows how the search is performed while shifting by two pixels.

[Method 3]: Tree search is performed. For example, a candidate for a certain search position is determined in advance, and only the vicinity of the position is searched.

[0013] One of the motion-compensated inter-frame prediction coding methods is to measure the motion of the entire screen such as panning of a camera, affine-transform the entire screen to create a reference image, and obtain a prediction error. There is a motion compensation method.
The parameter of the affine transformation at this time is called a global motion vector, and the process of obtaining the global motion vector is called global motion search. Video encoding system MPEG-
In No. 4, it is possible to combine global motion compensation and motion compensation that does not use affine transformation. For example,
Global motion compensation and motion compensation that does not use affine transformation can be selected for each macroblock. By selecting the one with the higher coding efficiency, the coding efficiency of the entire frame can be improved.

In the global motion search, there is a method of affine-transforming the entire screen to determine an affine transformation parameter that minimizes the prediction error of the entire screen, but there is also a method of using the result of the motion search for each macroblock. In this case, an affine transformation parameter is determined from the position of each macroblock in the screen and the motion vector, and is determined as a global motion vector. As the range of the motion search is increased, a global motion vector having a smaller prediction error is obtained. Conversely, as the range is increased, the amount of calculation increases.

[0015]

As described above, when the search range of the motion search is increased, the amount of calculation increases. Therefore, when the purpose is to encode a moving image in real time, if the amount of calculation increases, encoding cannot be performed in time, and a problem occurs in that encoded data at a target encoding speed cannot be obtained. In particular, when the purpose is to output coded data at a fixed frame rate, the fixed frame rate cannot be maintained if one frame cannot be coded in real time.

In the case where video encoding is executed by hardware, the operation processing speed in hardware is determined one-to-one for one operation instruction. Therefore, if the processing speed of the hardware is designed in advance so that the encoding can be executed even when the amount of calculation is maximized, the problem that the encoding cannot be completed does not occur.

On the other hand, when executing by software,
Since the arithmetic processing speed depends on the machine, the actual arithmetic processing speed is not determined one-to-one for one arithmetic instruction. Therefore, when it is executed on a machine with a low arithmetic processing speed, there is a problem that encoding cannot be performed in time. Other programs may be executed on the same machine, such as a personal computer (PC). In this case, if another program is executed during the moving image encoding, the arithmetic operation of the machine is devoted to the program, so that the encoding cannot be performed, and as a result, the encoding speed decreases. I do.
As described above, during the encoding, there is a problem that a time period during which the arithmetic processing for the encoding is not performed occurs.

[0018]

The arithmetic processing speed of a machine such as a personal computer is determined depending on the access speed of the CPU to the cache, the access speed to the memory, and the like. And the processing speed of the CPU executing the operation instruction. Further, since the CPU actually performs the operation, the number of instruction processings of the CPU for one operation instruction is the operation processing speed of the machine. That is, the result of dividing the number of instruction processings of the CPU for one operation instruction by the operating frequency of the CPU is the processing time of one operation instruction on the machine. Therefore, for example, in order to determine the processing time for encoding one frame, the number of instruction processings of the CPU during encoding is measured, and the result is divided by the operating frequency of the CPU.
Similarly, the processing time for executing the motion search processing can be obtained by measuring the number of instruction processings of the CPU during the execution of the motion search processing.

As described above, the processing time can be measured from the number of instruction processings of the CPU. Further, the number of instruction processes required for the encoding process that should be performed by the CPU depends only on the configuration of the program. Therefore, for example, in a case where the number of processing instructions necessary for encoding one frame is fixed, such as a motion search processing, it is possible to obtain the required number of instruction processing times before execution. In the encoding of the prediction error, the number of processing instructions differs because the number of DCT operations differs for each frame.

However, as mentioned in the above problem,
Other programs may be executed during the encoding. Since the CPU operates in a time-sharing manner, the CPU does not perform the encoding process while another program is being executed. Therefore, the instruction processing count of the CPU during encoding includes an instruction for executing another program. This causes the encoding speed to decrease when another program is executed during the encoding.

The present invention is characterized by changing the number of processing instructions required for a motion vector search or a global motion vector search by switching search accuracy or the like in order to solve the above problem. Specifically, the following method is used.

According to a first aspect of the present invention, in a moving picture coding method for performing a motion search by a computing unit and coding a moving picture using a motion compensation inter-frame predictive coding method, a predetermined frame before coding is determined. A coding target time which is a time for encoding one frame from the rate is set, and the number of motion search instructions which is the number of motion search processing instructions to be executed by the arithmetic unit with respect to the motion search method set in advance before coding. Information is stored, and the encoding time of each frame is measured during encoding.
After completion of frame encoding, a motion search target time for performing motion search of the next frame is set from the set encoding target time and the measured encoding time, and the stored motion search command is stored. According to the numerical information, a motion search method is determined from the motion search target time to control the motion search process.

Here, the motion search command number information may be information substantially corresponding to the number of commands, and may be, for example, information such as a value converted into time. The same applies to the inventions described below.

According to the first aspect of the present invention, the number of motion search processing instructions to be executed by the arithmetic unit is obtained for each motion search method set before encoding, so that the motion search time is set. Then, the motion search method can be determined by obtaining the number of processing instructions executed at that time.

For example, it is assumed that the following two types of motion search are set in advance.

[Method A]: The sum of absolute value differences is obtained for all pixels in the macroblock. The absolute value difference sum is obtained by shifting one pixel at a time within the search range, and the translation amount at the position where the absolute value difference sum is the smallest is defined as the motion vector.

[Method B]: Sum of absolute value differences is obtained for all pixels in the macroblock. The absolute value difference sum is obtained by shifting each pixel within the search range, and the translation amount at the position where the absolute value difference sum is the smallest is defined as the motion vector.

Before coding, the number of instructions to be executed by the arithmetic unit when motion search is performed by methods A and B is obtained in advance. As a result, it is assumed that method A has 10 instructions and method B has 5 instructions. After that, it is assumed that a motion search time is set when encoding is performed by an encoder having an arithmetic unit, and this is equivalent to seven instructions in the arithmetic unit. In this case, the method A cannot complete the motion search within seven instructions, but the method B can complete the motion search within seven instructions. Therefore, if the method B is selected and a motion search is performed, encoding can be performed in real time.

According to a second aspect of the present invention, in a moving picture coding method for performing a motion search by a computing unit and coding a moving picture by using a motion compensation inter-frame predictive coding method, a predetermined frame before coding is determined. A coding target time which is a time for encoding one frame from the rate is set, and the number of motion search instructions which is the number of motion search processing instructions to be executed by the arithmetic unit with respect to the motion search method set in advance before coding. Information is stored, and the encoding time of each frame is measured during encoding.
After completion of frame encoding, a motion search target time for performing motion search of the next frame is set from the set encoding target time and the measured encoding time, and the stored motion search command is stored. According to the numerical information, the motion search method is determined from the motion search target time, and the motion search process is controlled.
The execution time of the motion search process is measured, and the stored motion search command number information is updated from the actually measured motion search time.

According to the second invention, the number of motion search processing instructions set in advance in the first invention before encoding is reduced by the number of instructions actually processed by the encoder for executing the motion search. Can be updated. This method is suitable when it is not possible to determine the exact number of processing instructions before encoding. For example, when an encoder is configured by combining arithmetic units with different processing speeds, the overall processing speed of the encoder is not known, but the processing time required for motion search by actually coding several frames is not known. And the number of instructions. In addition, when the encoding process is performed by a general-purpose arithmetic unit such as a CPU, the processing time required for the actual encoding is not known because other programs may periodically use the CPU. However, according to the present invention, it is possible to obtain the processing time required for the motion search by actually performing encoding, and to obtain the number of instructions.

According to a third aspect of the present invention, in a moving picture coding method in which a motion search and a global motion search are performed by an arithmetic unit and a moving picture is coded by using a motion compensated inter-frame predictive coding method, 1 from the set frame rate
A coding target time, which is a time for coding a frame, is set, and information on the number of motion search instructions, which is the number of processing instructions for a motion search to be executed by a computing unit, is stored for a motion search method set before coding. In addition, information on the number of global motion search instructions, which is the number of processing instructions for motion search executed by the arithmetic unit, for the global motion search method set in advance before encoding is stored. , And after the completion of one frame encoding, the motion search target time for performing the motion search for the next frame and the global motion search from the set encoding target time and the measured encoding time. Is set, and a motion search method and a global motion search method are determined from the motion search target time according to the stored number of motion search command information and global motion search command information. Controlling the motion search and the global motion search.

According to the third aspect, when performing a global motion search in addition to the motion search, as in the first aspect, the motion search method and the motion search method can be executed within the target time. A global motion search method can be determined.

According to a fourth aspect of the present invention, in a moving picture coding method for performing a motion search and a global motion search by a computing unit and coding a moving picture using a motion compensation inter-frame predictive coding method, 1 from the set frame rate
A coding target time, which is a time for coding a frame, is set, and information on the number of motion search instructions, which is the number of processing instructions for a motion search to be executed by a computing unit, is stored for a motion search method set before coding. In addition, information on the number of global motion search instructions, which is the number of processing instructions for motion search executed by the arithmetic unit, for the global motion search method set in advance before encoding is stored. , And after the completion of one frame encoding, the motion search target time for performing the motion search for the next frame and the global motion search from the set encoding target time and the measured encoding time. Is set, and the motion search method and the global motion search method are determined from the motion search target time in accordance with the stored number of motion search commands and global motion search command information. Controlling the motion search process and the global motion search process, measuring the execution time of the motion search process, updating the stored motion search instruction number information from the actually measured motion search time, and The execution time of the global motion search process is measured, and the stored global motion search command number information is updated from the actually measured global motion search time.

According to the fourth aspect of the present invention, when performing a global motion search in addition to the motion search, similarly to the second aspect of the present invention, the processing time required for the motion search is obtained by actually coding the motion search. The number of instructions required for the search can be obtained.

According to a fifth aspect of the present invention, there is provided a moving picture coding method for performing a block motion search for each divided area of a screen by a computing unit and coding the moving picture using a motion compensation inter-frame predictive coding method. A coding target time, which is a time for coding one frame from a predetermined frame rate before coding, is set, and a motion search executed by an arithmetic unit is performed in accordance with a motion search method set before coding. The number of motion search instructions, which is the number of processing instructions, is stored, and the encoding time of each frame is measured during encoding. After the encoding of one frame is completed, the set encoding target time and the measured encoding time are measured. From the above, a motion search target time for performing a motion search for the next frame is set, a motion search time for each region executed in the block motion search process is measured, and the measured motion search for each block is performed. To determine the motion search method from the target motion search time and the frame motion search time in accordance with the stored motion search command count information to control the motion search process. I do.

According to the fifth aspect, a motion search is performed for each area to measure a motion search time, and the total value of the time can be compared with the motion search target time. this is,
This is suitable when the number of motion search instructions actually processed by the encoder during encoding is different from the number of motion search processing instructions set in advance before encoding. For example, when the total value of the motion search time becomes longer than the target time, the motion search of the frame can be forcibly terminated. As a region, a macroblock or a slice that is a set of macroblocks can be applied. 0 vector (stationary) in the area that could not be searched when forced termination
Alternatively, a motion vector at the same position in the previous frame or a motion vector of a surrounding area can be used as a motion vector of the area.

According to the sixth aspect of the present invention, a moving image code for performing a block motion search and a global motion search for each divided area of a screen by a computing unit, and coding the moving image using a motion compensation inter-frame predictive coding method. In the encoding method, a coding target time, which is a time for encoding one frame from a predetermined frame rate before encoding, is set, and the arithmetic unit executes the motion search method set in advance before encoding. The information on the number of motion search instructions, which is the number of motion search processing instructions to be executed, is stored, and the number of motion search processing instructions to be executed by the arithmetic unit is compared with the global motion search method set before encoding. The number of motion search instructions is stored,
At the time of encoding, the encoding time of each frame is measured, and after one frame encoding is completed, a motion search target for performing a motion search of the next frame is determined from the set encoding target time and the measured encoding time. Set the local motion search target time, which is the time, and the global motion search target time, which is the motion search target time for performing the global motion search, and measure the motion search time for each area executed in the block motion search process. The measured motion search time for each block is
Find the total frame motion search time for each frame,
The motion search time for each area executed in the global block motion search processing is measured, and the frame global motion search time obtained by summing up the measured motion search times for each block is calculated for each frame. A motion search method is determined from the local motion search target time and the frame motion search time in accordance with the search command number information to control the motion search process, and the global motion search is determined according to the stored global motion search command number information. A global motion search method is determined from the target time and the frame global motion search time, and the global motion search process is controlled.

According to the sixth aspect, when performing a global motion search in addition to the motion search, similar to the fifth aspect,
For example, when the number of motion search and global motion search instructions actually processed by the encoder during encoding is different from the number of motion search and global motion search processing instructions preset before encoding. When the total value of the motion search time and the global motion search becomes larger than the target time, the motion search and the global motion search for that frame can be forcibly terminated.

According to a seventh aspect of the present invention, there is provided a moving picture coding method for performing a block motion search for each divided area of a screen by a computing unit and coding the moving picture using a motion compensation inter-frame predictive coding method. A coding target time, which is a time for coding one frame from a predetermined frame rate before coding, is set, and a motion search executed by an arithmetic unit is performed in accordance with a motion search method set before coding. The number of motion search instructions, which is the number of processing instructions, is stored. Image statistical information of each area in the screen is detected from the image information of the input image, and the order of the area for motion search is determined from the detected image statistical information. The encoding time of one frame is determined, and after the encoding of one frame is completed, a motion search target for performing a motion search of the next frame is determined from the set encoding target time and the measured encoding time. A time is set, a motion search time for each area executed in the block motion search process is measured, and a frame motion search time obtained by summing the measured motion search times for each block in frame units is obtained and stored. A motion search method is determined from the target motion search time and the frame motion search time according to the information on the number of motion search commands, and the motion search process is controlled.

According to the seventh aspect, when the motion search is completed in the middle of the frame as in the fifth aspect, the order of the motion search for each area is determined before the motion search. It is possible to set. For example, before performing a motion search, it is possible to measure whether each area in the screen is a moving area and to perform a motion search process on the moving area before the still area. In the motion search of the area determined to be a static area before the motion search, the total value of the motion search time becomes larger than the target time, and even if the motion search for that frame is forcibly terminated, it is still. The motion vector of the region can be a zero vector. As a result, it is possible to reduce a decrease in coding efficiency from a case where a motion vector is actually obtained.

According to the eighth aspect of the present invention, a moving image code for performing a block motion search and a global motion search for each divided area of a screen by a computing unit, and coding the moving image using a motion compensation inter-frame predictive coding method. In the encoding method, a coding target time, which is a time for encoding one frame from a predetermined frame rate before encoding, is set, and the arithmetic unit executes the motion search method set in advance before encoding. The information on the number of motion search instructions, which is the number of motion search processing instructions to be executed, is stored, and the number of motion search processing instructions to be executed by the arithmetic unit is compared with the global motion search method set before encoding. The number of motion search instructions is stored,
Image statistical information of each area in the screen is detected from the image information of the input image, the order of the area for motion search is determined from the detected image statistical information, the encoding time of one frame is measured, and one frame is calculated. After the coding is completed, a local motion search target time, which is a motion search target time for performing a motion search for the next frame, and a global motion search are performed based on the set coding target time and the measured coding time. The global motion search target time, which is the target time for motion search, is set, the motion search time for each region executed in the block motion search process is measured, and the measured motion search time for each block is totaled for each frame. Find the frame motion search time, measure the motion search time for each area executed in the global block motion search process, and search for the measured motion for each block. A frame global motion search time is calculated by summing the time intervals in frame units, and a motion search method is determined from the local motion search target time and the frame motion search time in accordance with the stored motion search command number information. , And a global motion search method is determined from the global motion search target time and the frame global motion search time according to the stored global motion search command number information, and the global motion search process is controlled.

According to the eighth aspect, similarly to the seventh aspect, the motion search is forcibly performed in the middle of the frame by setting the order of the motion search for each area before performing the motion search. It is possible to reduce a decrease in encoding efficiency at the time of termination.
In addition, before performing a motion search, it is determined in advance whether or not to perform a motion search for each region, and during the encoding, a motion search is performed only in that region, and the time allocated to the global motion search is increased. It is possible. For example, before performing a motion search, it is determined in advance whether each region is a moving region or a still region, and only the moving region is searched for a motion, and the still region is not searched for a motion. As a result, the time for the motion search in the still area can be allocated to the global motion search.

A software program for realizing the above-described processing method by a computer can be stored in an appropriate recording medium such as a computer-readable portable medium memory, a semiconductor memory, and a hard disk.

[0044]

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

In the first embodiment, before encoding, the number of processing instructions for each motion search method and the number of processing instructions for each global motion search method are set in advance. Is described below.

On a computer equipped with a CPU having an operating frequency of 500 MHz, the screen size is CIF (352 horizontal).
The operation when encoding a moving image of (pixels, 288 pixels vertically) at a frame rate of 10 frames per second will be described.

As a motion search method, the following methods A, B and C are prepared. Apply the same method to all macroblocks in the screen.

[Method A]: Sum of absolute value differences is obtained for all pixels in a macroblock. The search range is 16 pixels vertically.
The width is 16 pixels. The absolute value difference sum is obtained by shifting one pixel at a time within the search range, and the translation amount at the position where the absolute value difference sum is the smallest is defined as the motion vector.

[Method B]: Sum of absolute value differences is obtained for all pixels in a macroblock. The search range is 16 pixels vertically.
The width is 16 pixels. The absolute value difference sum is obtained by shifting each pixel within the search range, and the translation amount at the position where the absolute value difference sum is the smallest is defined as the motion vector.

[Method C]: The motion vectors of all macroblocks are set to 0 vectors.

Assume that the number of instructions for the motion search in the method A is 10 million, the number of instructions for the motion search in the method B is 5 million, and the number of instructions for the method C is 10,000.

As a global motion search method,
First, in each case, the screen is divided into blocks of 8 pixels vertically and 8 pixels horizontally, and motion search is performed for each block. After the motion search for all blocks is completed, the number of blocks for the motion vector is totaled, and the motion vector with the largest number of blocks is set as a global motion vector. For the motion search in block units, the following methods D, E and F are prepared. Apply the same method to all blocks in the screen.

[Method D]: Sum of absolute value differences is obtained for all pixels in the block. The search range is 32 pixels vertically and 32 pixels horizontally
Pixels. The absolute value difference sum is obtained by shifting one pixel at a time within the search range, and the translation amount at the position where the absolute value difference sum is the smallest is defined as the motion vector.

[Method E]: The sum of absolute value differences is obtained for all pixels in the block. The search range is 32 pixels vertically and 32 pixels horizontally
Pixels. The absolute value difference sum is obtained even by two pixels within the search range, and the parallel movement amount at the position where the absolute value difference sum is the smallest is defined as the motion vector.

[Method F]: The motion vectors of all blocks are set to 0 vectors.

The number of instructions in the global motion search when the method D is adopted is 30 million, the number of instructions in the global motion search when the method E is adopted is 15 million, and the number of instructions in the global motion search when the method F is adopted. Assume that the number of instructions is 10,000.

FIG. 4 shows a processing procedure of the encoding method in the first embodiment. Note that “parts” described in each block shown in FIG. 4 can be read as “procedures”. The same applies to other embodiments.

The motion search control unit (motion search control procedure) 106 shown in FIG. 4 selects one of the preset methods A, B, and C, and the method D, the method E, or the method F. Select one from

FIG. 5 shows the flow of selection. The motion search target time is set to target_me_time, and the value obtained by converting the number of instructions from method A to method F into time is me_t
im [i] (i = 1: method A, i = 2: method B, i =
3: Method C, 1 = 4: Method D, i = 5: Method E, i =
6: Method F). Make the following selections:

Me_time [1] and me_time
The sum of [4] is the motion search target time (target_me
If it is smaller than _time), the method A and the method D are selected (step S10).

Me_time [1] and me_time
The sum of [5] is the motion search target time (target_me
If it is smaller than _time), method A and method E are selected (step S11).

Me_time [1] and me_time
The sum of [6] is the motion search target time (target_me
If it is smaller than _time), method A and method F are selected (step S12).

Me_time [2] and me_time
The sum of [6] is the motion search target time (target_me
If it is smaller than _time), the method B and the method F are selected (step S13).

Me_time [3] and me_time
The sum of [6] is the motion search target time (target_me
If it is smaller than _time), the method C and the method F are selected (step S14).

As described above, no matter which combination is selected, the selection is made so that the motion search and the global motion search are processed within the motion search target time. If it is not within the motion search target time when the methods C and F are selected (NO in step S14), the encoding of the next frame is started without encoding the frame. In this case, an encoding target time for encoding one frame is added to the motion search target time.

First, before starting the encoding, the target time setting unit 10
In step 1, since the frame rate is 10 frames per second, 100 msec is set as the target encoding time. The motion parameter storage unit 102 stores the number of processing instructions from method A to method C, and the global motion parameter storage unit 103 stores the number of processing instructions from method D to method F.

In this embodiment, the actual time corresponding to the number of processing instructions is as follows.

Me_time [1] = 10 million / 500 MHz * 1000 msec = 20 msec me_time [2] = 5 million / 500 MHz * 1000 msec = 10 msec me_time [3] = 10,000 / 500 MHz * 1000 msec = 0.02 msec me_time [4] = 3000 10,000 / 500 MHz * 1000 msec = 60 msec me_time [5] = 15 million / 500 MHz * 1000 msec = 30 msec me_time [6] = 10,000 / 500 MHz * 1000 msec = 0.02 msec Subsequently, the J-th and J + 1-th after coding is started A frame encoding procedure will be described. It is assumed that the encoding is not in time for the encoding target time in the J-1st frame, and the encoding is in time for the encoding target time in the Jth frame. It is assumed that the coding time measured by the coding time measuring unit 104 is 1 using the method A for the motion search and the method D for the global motion search for the J-1st frame.
It is assumed that the time is 10 msec. That is, 10 msec
The delay has occurred.

In this situation, the motion search time setting section 105 first obtains a motion search target time for the J-th frame. Using method A and method D, the encoding time is 1
Since it was 10 msec, the processing time other than the motion search can be calculated as 110 msec− (method A time + method D time) = 110 msec− (20 msec + 60 msec) = 30 msec.

Considering the recovery from the delay of 10 msec in the next frame, the motion search target time target_
me_time is set as 100 msec− (delay time) − (time other than motion search) = 100 msec−10 msec−30 msec = 60 msec. The motion search control unit 106 selects the method A and the method E according to the selection flow of FIG.

The image input unit 107 acquires one frame of image information. The motion search unit 108 performs a motion search using the method A, and the global motion search unit 109 performs a global motion search using the method E. The inter-frame prediction encoding unit 110 encodes the image information of the J-th frame using the motion vector obtained by the motion search unit 108 and the global motion vector obtained by the global motion search unit 109. The local decoding unit 111 decodes the encoded data. The encoding time measuring unit 104 measures the encoding time of the J-th frame. The encoding time is 55 msec
Assume that

In this case, the delay of 10 msec generated in the J-1st frame is eliminated, and the encoding is performed in real time. As described above, even when it is not possible to perform encoding in real time in the previous frame, the delay can be eliminated in the next frame by adjusting the motion search time of the next frame.

Subsequently, in the motion search time setting unit 105,
A motion search target time for the (J + 1) th frame is obtained. 55 msec coding time using method A and method E
Therefore, the time of the processing other than the motion search can be calculated as 55 msec− (time of method A + time of method E) = 50 msec− (20 msec + 30 msec) = 5 msec. Therefore, the motion search target time target
_Me_time is 100 ms- (time other than motion search) = 100 ms
ec−5 msec = 95 msec. The motion search control unit 106 selects the method A and the method D according to the selection flow of FIG.

The image input unit 107 acquires one frame of image information. The motion search unit 108 performs a motion search using the method A, and the global motion search unit 109 performs a global motion search using the method D. The inter-frame prediction encoding unit 110 encodes the image information of the J-th frame using the motion vector obtained by the motion search unit 108 and the global motion vector obtained by the global motion search unit 109. The local decoding unit 111 decodes the encoded data. The encoding time measuring unit 104 measures the encoding time of the (J + 1) th frame.

As described above, when the encoding time of the previous frame is within the encoding target time, it is possible to increase the encoding efficiency over the previous frame by increasing the motion search processing time of the next frame. It is possible.

As in the first embodiment, before encoding,
The number of processing instructions for each motion search method and the number of processing instructions for each global motion search method are set, and moving images are encoded in real time by changing the motion search processing and global motion search processing according to these values. it can.

In the first embodiment, as a method for changing the number of processing instructions for the motion vector search and the global motion search, a method of shifting one pixel at a time within the search range (method A and method D), and a method of changing the number of processing instructions at the search range. The method of shifting two pixels at a time (methods B and E) and the method of calculating only when the motion vector is a zero vector (methods C and F) have been described. However, the number of processing instructions can be reduced by shifting the number of pixels by three or more pixels. Can be further reduced. Also, only the case where the search range of the motion vector is 16 pixels vertically and 16 pixels horizontally and the search range of the global motion search is 32 pixels vertically and 32 pixels horizontally, the number of processing instructions is reduced by expanding or reducing the search range. It can be changed. The second to fourth shown below
The same applies to the embodiment.

When the global motion search is not used, the procedure in which the processing of the global motion parameter storage section 103 and the processing of the global motion search section 109 are omitted in the first embodiment shown in FIG. Needless to explain. The same applies to other embodiments described below.

In the second embodiment, the number of processing instructions for each motion search method and the number of processing instructions for each global motion search method are set before encoding, and the moving picture is encoded according to these values. A mechanism for changing the number of processing instructions during encoding will be described.

On a computer equipped with a CPU with an operating frequency of 500 MHz, the screen size is CIF (352 horizontal).
The operation when encoding a moving image of (pixels, 288 pixels vertically) at a frame rate of 10 frames per second will be described.

The method A, the method B, and the method C are prepared as in the first embodiment, and the global motion search method is the method D, the method E, and the method F as in the first embodiment. Prepare The same motion search method and global motion search method are applied to all macroblocks in a screen.

FIG. 6 shows a processing procedure of an encoding method according to the second embodiment. Here, the selection flow of the motion prediction method and the global motion search method in the motion search control unit 206 is the same as that in the first embodiment shown in FIG.

The encoding procedure of the J-th and J + 1-th frames after starting the encoding will be described. In the J-th frame, the encoding is not in time for the target encoding time, and in the J-th frame, the encoding is in time for the target encoding time. As a premise, it is assumed that the coding time measured by the coding time measuring unit 204 is 120 msec using the method A for the motion search and the method D for the global motion search for the J-1st frame. That is, a delay of 20 msec occurs.

It is assumed that the motion search time measured by the motion search time measuring section 212 is 40 msec, and the global motion search time measured by the global motion search time measuring section 214 is 60 msec.

The number of instructions of each method stored in the motion parameter storage unit 202 is 10 million for the method A.
It is assumed that method B is 5 million and method C is 10,000. The number of instructions of each method stored in the global motion parameter storage unit 203 is 30 million for method D and 15 for method E.
It is assumed that the method F is 10,000 and the method F is 10,000.

In this situation, first, the number of motion search commands and the number of global motion search commands for the J-th frame are updated. Method A in the motion parameter updating unit 213
Is updated to 40 msec * 500 MHz = 20 million, and the global motion parameter updating unit 215 updates the number of instructions of method D to 60 msec * 500 MHz = 30 million.

Therefore, the actual time corresponding to the number of processing instructions is as follows.

Me_time [1] = 20 million / 500 MHz * 1000 msec = 40 msec me_time [2] = 5 million / 500 MHz * 1000 msec = 10 msec me_time [3] = 10,000 / 500 MHz * 1000 msec = 0.02 msec me_time [4] = 3000 10,000 / 500 MHz * 1000 msec = 60 msec me_time [5] = 15 million / 500 MHz * 1000 msec = 30 msec me_time [6] = 10,000 / 500 MHz * 1000 msec = 0.02 msec Subsequently, the motion search time setting unit 205 sets the J-th frame. To find the motion search target time for. Method A and Method D
Since the encoding time was 120 msec using, the processing time other than the motion search can be calculated as follows: 120 msec− (time of method A + time of method D) = 120 msec− (40 msec + 60 msec) = 20 msec.

In consideration of the recovery from the delay of 20 msec in the next frame, the motion search target time target_
me_time is set as 100 msec− (delay time) − (time other than motion search) = 100 msec−20 msec−20 msec = 60 msec. The motion search control unit 206 selects the method A and the method F according to the flow of FIG.

The image input unit 207 acquires one frame of image information. The motion search unit 208 performs a motion search using the method A, and the global motion search unit 209 performs a global motion search using the method F. Inter-frame prediction encoding unit 2
In step 10, the image information of the J-th frame is encoded using the motion vector obtained by the motion search unit 208 and the global motion vector obtained by the global motion search unit 209. The local decoding unit 211 decodes the encoded data. The encoding time measuring unit 204 measures the encoding time of the J-th frame. It is assumed that the encoding time is 55 msec. Also, it is assumed that the motion search time measured by the motion search time measurement unit 212 is 40 msec, and the global motion search time measured by the global motion search time measurement unit 214 is 0.1 msec.

In this situation, first, the number of motion search commands and the number of global motion search commands for the (J + 1) th frame are updated. In the motion parameter updating unit 213, the number of instructions of the method A is set to 40 msec * 500 MHz = 2000.
The number of instructions of the method F is updated to 0.1 msec * 500 MHz = 50,000 by the global motion parameter updating unit 215.

Therefore, the actual time corresponding to the number of processing instructions is as follows.

Me_time [1] = 20 million / 500 MHz * 1000 msec = 40 msec me_time [2] = 5 million / 500 MHz * 1000 msec = 10 msec me_time [3] = 10,000 / 500 MHz * 1000 msec = 0.02 msec me_time [4] = 3000 10,000 / 500 MHz * 1000 msec = 60 msec me_time [5] = 15 million / 500 MHz * 1000 msec = 30 msec me_time [6] = 50,000 / 500 MHz * 1000 msec = 0.1 msec Subsequently, the motion search time setting unit 205 sets the J + 1-th frame. To find the motion search target time for. Since the encoding time was 55 msec using the method A and the method F, the processing time other than the motion search was 55 msec− (the time of the method A + the time of the method E) = 55 msec− (40 msec + 0.1 msec) = It can be calculated as 14.9 msec. Therefore, the motion search target time target_me_time is set as 100 msec− (time other than motion search) = 100 msec−14.9 msec = 85.1 msec. The motion search control unit 206 selects the method A and the method E according to the flow of FIG.

An image input unit 207 acquires image information for one frame. The motion search unit 208 performs a motion search using the method A, and the global motion search unit 209 performs a global motion search using the method E. The inter-frame predictive coding unit 210 codes the image information of the (J + 1) th frame using the motion vector obtained by the motion search unit 208 and the global motion vector obtained by the global motion search unit 209. The local decoding unit 211 decodes the encoded data. The encoding time measuring unit 204 measures the encoding time of the (J + 1) th frame.

As described above, in the second embodiment, the number of motion search instructions and the number of global motion search instructions in the arithmetic unit are obtained and updated from the encoding of the previous frame, thereby obtaining
It is possible to select a motion search method and a global motion search method according to the arithmetic processing speed of the arithmetic unit.

In the second embodiment, the motion search unit 2
08, the motion search time measuring unit 212
It is also preferable to measure the motion search time by using, and use this time to reset the motion search target time when performing the global motion search again. For example, assuming that the motion search time in the (J + 1) th frame is 20 msec, the motion search target time for the global motion search in the (J + 1) th frame is set as target_me_time = 85.1 msec-20 msec = 65.1 msec. The global motion search method is determined again based on this information. FIG. 7 shows this selection flow. Method D is selected according to this selection flow (step S2
0), global motion search is performed. The operation after obtaining the global motion vector is the same as the processing procedure of FIG.

In the selection flow of FIG. 7, target_m
If e_time is greater than me_time [4] (step S20), method D is selected and target_
me_time is less than or equal to me_time [4] and m
If it is larger than e_time [5] (step S2
1), Method E is selected. Furthermore, target_me
_Time is less than or equal to me_time [5] and me_
If it is larger than time [6] (step S2
2), Method F is selected, and if it is less than me_time [6], global motion compensation is not performed.

As described above, by resetting the motion search target time for the global motion search after the motion search, a global motion vector with a smaller prediction error can be obtained.

In the third embodiment, a description will be given of a mechanism in which the number of processing instructions for each motion search method is set in advance before encoding, and a moving image is encoded in real time according to this value. However, unlike the first embodiment, the motion search is controlled by measuring the motion search time in macroblock units. Prior to encoding, the priority of motion search is determined for each macroblock in advance. The following example shows an operation example of forcibly terminating when a motion search cannot be performed on all macroblocks within the motion search target time.

On a computer equipped with a CPU having an operating frequency of 500 MHz, the screen size is CIF (352 horizontal).
The operation when encoding a moving image of (pixels, 288 pixels vertically) at a frame rate of 10 frames per second will be described.

As a motion search method, a method A and a method B described in the first embodiment are prepared. It is assumed that the number of instructions for the motion search in the method A is 10 million and the number of instructions for the motion search in the method B is 5 million. However, different from the first embodiment, the motion search method is appropriately changed for a macroblock in a screen. FIG. 8 shows a processing procedure of the encoding method in the third embodiment. FIG. 9 shows a flow of selecting a motion prediction method performed by the motion search control unit 306.

The motion search time for each block measured by the block motion search time measuring unit 312 is totaled for each frame, and the total frame motion search time is calculated as total.
_Me_time, and target motion search time
t_me_time, and the value obtained by converting the number of instructions up to method A and method B into time is me_time [i] (i =
1: Method A, i = 2: Method B). The number of macroblocks count_MB for which the motion search has been performed is measured by counting the number of times the motion search process has been performed. It is assumed that the number of macro blocks in the screen is total_MB.

Further, the block priority detection section 316
Edge pixels in the screen are detected, and the number of edge pixels is counted for each macroblock. Since edges are important for maintaining visual quality, they are important for improving image quality. Therefore, it is necessary to improve the coding efficiency of a macroblock including many edge pixels. The search block determination unit 317 creates a list in which the numbers of the macroblocks are arranged in the descending order of the number of edge pixels.

The operation of the present embodiment in such a procedure configuration will be described. First, before starting the encoding, the target time setting unit 30
In step 1, since the frame rate is 10 frames per second, 100 msec is set as the target encoding time. Further, the motion parameter storage unit 302 stores the number of processing instructions of the method A and the method B.

Next, the encoding procedure of the J-th frame after the start of encoding will be described. As a premise, it is assumed that the encoding time measured by the encoding time measuring unit 304 in the J-1st frame is 90 msec. Also,
It is assumed that total_me_time is 80 msec.

In this situation, first, the motion search time setting section 305 obtains a motion search target time for the J-th frame. Since the encoding time is 90 msec, the processing time other than the motion search can be calculated as 90 msec−80 msec = 10 msec. Motion search target time ta in the next frame
rget_me_time is set as 100 msec− (time other than motion search) = 100 msec−10 msec = 90 msec. Therefore, the motion search method A is selected for the first macroblock.

The image input unit 307 acquires one frame of image information. The block priority detection unit 316 performs edge detection, and counts the number of edge pixels for each macroblock. The search block determination unit 317 creates a list in which the macroblock numbers are arranged in the order of the macroblocks with more edge pixels. FIG. 10 shows an example of the list created here. The block motion search unit 308 determines a motion search method according to the selection flow in FIG. 9 and performs a motion search in macroblock units in the order of the list in FIG.

In the selection of the motion method, as shown in FIG.
The difference between total_MB and count_MB is calculated as total
_MB divided by me_time [1] is
target_me_time and total_me_t
If the difference is smaller than the difference (step S30), the method A is selected. Otherwise, calculate the product of the difference between total_MB and count_MB divided by total_MB and me_time [2], which is
target_me_time and total_me_ti
If it is smaller than the difference from me (step S31), method B
Select In other cases, the motion search is stopped.

After encoding the macro block of macro block number 200, the frame motion search time is set to 69.
If it is 9999999 msec, the motion search is not performed on the macroblocks after the 200th among the macroblocks listed in the list of FIG. 10 according to the selection flow of FIG. The motion vector of a macroblock for which no motion search is performed is 0 vector.

The inter-frame predictive coding unit 310 codes the image information of the J-th frame using the motion vector obtained by the block motion search unit 308. The local decoding unit 311 decodes the encoded data. The encoding time measuring unit 304 measures the encoding time of the J-th frame.

As described above, it is possible to end the motion search processing in the middle of the frame by measuring the motion search processing time for each macroblock. Even if the motion search is terminated in the middle of a frame, the coding efficiency of a macroblock containing many edge pixels is improved by determining the order of motion search for each macroblock before coding. However, it is possible to reduce the deterioration of the image quality.

In the fourth embodiment, before encoding, the number of processing instructions for each motion search method and the number of processing instructions for each global motion search method are set in advance, and moving images are recorded in real time according to these values. A mechanism for encoding an image will be described.
However, different from the first embodiment, the motion search time is measured in macroblock units to control the motion search, and the global motion search time is measured in macroblock units to control the global motion search. Prior to encoding, the priorities of motion search and global motion search are determined for each macroblock in advance. In the following example, an operation example in which the block motion search unit 408 forcibly terminates when a motion search cannot be executed for all macroblocks within the local motion search target time.

On a computer equipped with a CPU having an operating frequency of 500 MHz, the screen size is CIF (352 horizontal).
The operation when encoding a moving image of (pixels, 288 pixels vertically) at a frame rate of 10 frames per second will be described.

For the motion search method, methods A and B are prepared as in the first embodiment, and for the global motion search method, methods D and E are prepared as in the first embodiment.
It is assumed that the number of instructions for the motion search in the method A is 10 million and the number of instructions for the motion search in the method B is 5 million. It is assumed that the number of instructions in the global motion search when the method D is adopted is 30 million, and the number of instructions in the global motion search when the method E is adopted is 15 million. However, different from the first embodiment, the motion search method and the global motion search method are appropriately changed for a macroblock in a screen.

FIG. 11 shows a processing procedure of an encoding method according to the fourth embodiment. In the motion search time setting unit 405,
From the motion search target time target_me_time to the local motion search target time target_local_m
e_time and global motion search target time target
Determine t_global_me_time. Method A
And the value obtained by converting the number of instructions in method B into time is me_tim
e [i] (i = 1: method A, i = 2: method B), and the value obtained by converting the number of instructions of method D and method E into time is me_t
It is assumed that im [i] (i = 4: method D, i = 5: method E). It is calculated according to the following equation.

Target_local_me_time = target_me_time * me_time
[1] / (me_time [l] + me_time
[4]) target_global_me_time = target_me_time-target_lo
cal_me_time FIG. 12 shows a flow of selecting a motion prediction method in the motion search control unit 406, and FIG. 13 shows a flow of selecting a global motion prediction method.

The motion search time for each block measured by the block motion search time measuring unit 412 is totaled for each frame, and the total frame motion search time is calculated as total.
_Local_me_time. The number of macroblocks count_MB for which the motion search has been performed is measured by counting the number of times the motion search process has been performed. It is assumed that the number of macro blocks in the screen is total_MB.

Global block motion search time measuring section 4
The frame global motion search time obtained by summing the motion search time for each block measured at step 19 in frame units is t
Let it be total_global_me_time.

The image information of the pixels in each frame is 0 to
The block priority detection unit 416 calculates the absolute value of the difference between pixels at the same position between the original image of the current frame and the locally decoded image of the previous frame, and determines the absolute value of the difference. The number of pixels of 50 or more (here, called moving pixels) is measured for each macroblock. As the number of video elements increases, the macroblock can be determined as a moving area.
The search block determination unit 417 creates a list in which the numbers of the macroblocks are arranged in the descending order of the number of moving pixels.

The operation of the present embodiment in such a procedure configuration will be described. First, a target time setting unit 40 before encoding starts.
In step 1, since the frame rate is 10 frames per second, 100 msec is set as the target encoding time. Further, the motion parameter storage unit 402 stores the number of processing instructions of the method A and the method B.

Next, the encoding procedure of the J-th frame after the start of encoding will be described. It is assumed that the encoding time measured by the encoding time measuring unit 404 in the J-1st frame is 100 msec. The total_local_me_time is 20 m
sec and total_g1obal_me_ti
It is assumed that me is 50 msec.

In this situation, first, the motion search time setting section 405 obtains a motion search target time for the J-th frame. Since the encoding time was 100 msec, the processing time other than the motion search is 100 msec−20 msec−50 msec = 30 m
sec. Motion search target time ta in the next frame
rget_me_time is set as follows: 100 msec− (time other than motion search) = 100 msec−30 msec = 70 msec. From now on, the local motion search target time ta
rget_local_me_time and global motion search target time target_global_me_t
and im.

Target_local_me_time = 70 msec * 20 msec / (20 msec + 60 msec) = 17.5 msec target_global_me_time = 70 msec−17.5 msec = 52.5 msec Image input unit 407 acquires one frame of image information. The block priority detection unit 416 calculates the absolute value difference between the original image of the current frame and the locally decoded image of the previous frame, and counts the number of moving pixels for each macroblock.
The search block determination unit 417 creates a list in which the macroblock numbers are arranged in the order of the macroblock having the largest number of moving pixels. The list created here is the same as in FIG.

The block motion search section 408 determines a motion search method according to the selection flow in FIG. 12, and performs a motion search in macroblock units in the order of the list in FIG. After encoding the macroblock of macroblock number 100, the frame motion search time is 17.499999999.
If msec is reached, the motion search is not performed on the macroblocks after the 100th among the macroblocks listed in the list of FIG. 10 according to the selection flow of FIG. The motion vector of a macroblock for which no motion search is performed is 0 vector.

The global block motion search section 409 determines a global motion search method according to the selection flow in FIG. 13, and performs a motion search in macroblock units in the order of the list in FIG. Assuming that the frame global motion search time becomes 52.499999999 msec after encoding the macroblock of macroblock number 50, among the macroblocks listed in the list of FIG. 10 according to the selection flow of FIG. No motion search is performed for macroblocks after the number. The motion vector of a macroblock for which no motion search is performed is 0 vector. Of these motion vectors in units of macroblocks, the motion vector having the largest number of macroblocks is defined as a global motion vector.

The inter-frame predictive coding unit 410 codes the image information of the J-th frame using the motion vector obtained by the block motion search unit 408 and the global motion vector obtained by the global block motion search unit 409. Become The local decoding unit 411 decodes the encoded data. The encoding time measuring unit 404 measures the encoding time of the J-th frame.

As described above, the local motion search target time and the global motion search target time are set from the motion search target time, and the motion search processing time and the global motion search processing time are measured for each macroblock. It is possible to end the motion search processing and the global motion search processing time.

When the block motion search section 408 finishes within a preset local motion search target time for all macroblocks, it is preferable to add the remaining time to the global motion search target time. It is. For example, if the frame motion search time measured by the block motion search time measuring unit 412 is 9 msec and the block motion search unit 408 is finished, the global motion search target time target_global_me_time is set to 52.5 msec + (17.5 msec-9 msec).
= 61 msec. According to this value, the global motion search method is determined according to the selection flow of FIG. In this case, method D is selected for the first macroblock. Thereafter, encoding is performed according to the procedure shown in FIG.

As described above, when the block motion search processing is completed earlier than the target, a global motion vector having a smaller prediction error can be obtained by determining the global motion search target time again.

The selection algorithm shown in the selection flow of the motion search shown in FIG. 12 and the selection flow of the global motion search shown in FIG. 13 is the same as the example shown in FIG.
Detailed description for each step is omitted.

[0131]

According to the present invention, the number of processing instructions for each motion search method and the number of processing instructions for each global motion search method are set before encoding, and the motion search processing and global By changing the motion search processing, a moving image can be encoded in real time. That is, as in the case of encoding with a personal computer, the encoding process is stopped while another program is being executed, or the processing speed of the CPU differs depending on the model. Even in a different case, the time required for the motion vector search can be adjusted, and the moving image can be encoded in real time.

[Brief description of the drawings]

FIG. 1 is a diagram illustrating a relationship between a macroblock and a search range.

FIG. 2 is a diagram illustrating an example of a pixel selection method.

FIG. 3 is a diagram illustrating an example in which a search position is shifted by two pixels.

FIG. 4 is a diagram illustrating the procedure of the first embodiment.

FIG. 5 is a flowchart illustrating a motion search selection flow according to the first embodiment;

FIG. 6 is a diagram illustrating a procedure of the second embodiment.

FIG. 7 is a selection flow of a global motion search in the second embodiment.

FIG. 8 is a diagram for explaining the procedure of the third embodiment.

FIG. 9 is a flowchart illustrating a motion search selection flow according to the third embodiment.

FIG. 10 is a diagram illustrating an example of the order of macroblocks for which motion search is performed.

FIG. 11 is a diagram illustrating a procedure according to a fourth embodiment.

FIG. 12 is a flowchart illustrating a motion search selection flow according to a fourth embodiment.

FIG. 13 is a selection flow of a global motion search in the fourth embodiment.

[Explanation of symbols]

 Reference Signs List 101 Target time setting unit 102 Motion parameter storage unit 103 Global motion parameter storage unit 104 Encoding time measurement unit 105 Motion search time setting unit 106 Motion search control unit 107 Image input unit 108 Motion search unit 109 Global motion search unit 110 Inter-frame prediction Encoding unit 111 Local decoding unit

 ──────────────────────────────────────────────────続 き Continued on the front page F-term (reference)

Claims (9)

[Claims] An image input step of inputting an image and a motion search are performed in a moving image coding method for performing a motion search by an arithmetic unit and coding the moving image using a motion compensation inter-frame predictive coding method. A moving image that executes a motion search step, an inter-frame predictive encoding step of performing motion-compensated inter-frame predictive encoding using the motion vector obtained in the motion search step, and a local decoding step of decoding encoded data An encoding method, comprising: a target time setting step of setting an encoding target time which is a time for encoding one frame from a predetermined frame rate before encoding; and a motion search method preset before encoding. A motion parameter storing step of storing information on the number of motion search instructions, which is the number of processing instructions for motion search executed by the arithmetic unit, and encoding of one frame From the encoding time measuring step for measuring the time, and the encoding target time set in the target time setting step and the encoding time measured in the encoding time measuring step after one frame encoding is completed, According to a motion search time setting step of setting a motion search target time for performing a motion search of a frame, and motion search command number information stored in a motion parameter storage step,
A motion search control step of determining a motion search method from a motion search target time and controlling a motion search process; and performing a motion search adaptive image encoding method. 2. A moving picture coding method for performing a motion search by an arithmetic unit and coding a moving picture by using a motion compensation inter-frame predictive coding method, wherein an image input step of inputting an image and a motion search are performed. A moving image that executes a motion search step, an inter-frame predictive encoding step of performing motion-compensated inter-frame predictive encoding using the motion vector obtained in the motion search step, and a local decoding step of decoding encoded data An encoding method, comprising: a target time setting step of setting an encoding target time which is a time for encoding one frame from a predetermined frame rate before encoding; and a motion search method preset before encoding. A motion parameter storing step of storing information on the number of motion search instructions, which is the number of processing instructions for motion search executed by the arithmetic unit, and encoding of one frame From the encoding time measuring step for measuring the time, and the encoding target time set in the target time setting step and the encoding time measured in the encoding time measuring step after one frame encoding is completed, According to a motion search time setting step of setting a motion search target time for performing a motion search of a frame, and motion search command number information stored in a motion parameter storage step,
A motion search control step for determining a motion search method from the motion search target time and controlling the motion search process; a motion search time measuring step for measuring the execution time of the motion search step;
A motion parameter updating step of updating the motion search command number information stored in the motion parameter storage step from the motion search time measured in the motion search time measuring step; Encoding method. 3. A moving picture coding method for performing a motion search and a global motion search in an arithmetic unit and coding a moving picture using a motion compensation inter-frame predictive coding method, comprising: A motion search step for performing a motion search, a global motion search step for performing a global motion search, and a motion compensation inter-frame prediction code using the motion vector obtained in the motion search step and the global motion vector obtained in the global motion search step. A moving picture coding method for performing an inter-frame predictive coding step of performing coding and a local decoding step of decoding coded data, wherein one frame is coded from a predetermined frame rate before coding. Target time setting step for setting an encoding target time, which is the time to perform, and a motion search method preset before encoding. A motion parameter storage step for storing information on the number of motion search instructions, which is the number of motion search processing instructions executed by the arithmetic unit, and a global motion search method set in advance before encoding. A global motion parameter storage step for storing information on the number of global motion search instructions that are the number of motion search processing instructions to be executed by the encoder, an encoding time measuring step for measuring the encoding time of one frame, and one frame encoding After the end, from the encoding target time set in the target time setting step and the encoding time measured in the encoding time measuring step,
A motion search time setting step for setting a motion search target time for performing a motion search for the next frame and a global motion search; and information on the number of motion search instructions stored in the motion parameter storage step from the motion search target time. A motion search method and a global motion search method are determined according to the global motion search command number information stored in the global motion parameter storage step, and a motion search control step for controlling the motion search process and the global motion search process is executed. Processing time adaptive image encoding method. 4. A moving picture coding method for performing a motion search and a global motion search in an arithmetic unit and coding a moving picture using a motion compensated inter-frame predictive coding method, comprising the steps of: A motion search step for performing a motion search, a global motion search step for performing a global motion search, and a motion compensation inter-frame prediction code using the motion vector obtained in the motion search step and the global motion vector obtained in the global motion search step. A moving picture coding method for performing an inter-frame predictive coding step of performing coding and a local decoding step of decoding coded data, wherein one frame is coded from a predetermined frame rate before coding. Target time setting step for setting an encoding target time, which is the time to perform, and a motion search method preset before encoding. A motion parameter storage step for storing information on the number of motion search instructions, which is the number of motion search processing instructions executed by the arithmetic unit, and a global motion search method set in advance before encoding. A global motion parameter storage step for storing information on the number of global motion search instructions that are the number of motion search processing instructions to be executed by the encoder, an encoding time measuring step for measuring the encoding time of one frame, and one frame encoding After the end, from the encoding target time set in the target time setting step and the encoding time measured in the encoding time measuring step,
A motion search time setting step for setting a motion search target time for performing a motion search for the next frame and a global motion search; and information on the number of motion search instructions stored in the motion parameter storage step from the motion search target time. A motion search control step for determining a motion search method and a global motion search method in accordance with the global motion search command number information stored in the global motion parameter storage step, and controlling the motion search process and the global motion search process; A motion search time measurement step for measuring the execution time of the step, and a motion parameter update step for updating the information of the number of motion search instructions stored in the motion parameter storage step from the motion search time measured in the motion search time measurement step And a global measure the execution time of the global motion search step. And Bal motion search time measurement step,
Performing a global motion parameter updating step of updating information on the number of global motion search commands stored in the global motion parameter storage step from the global motion search time measured in the global motion search time measuring step. Processing time adaptive image coding method. 5. A moving picture coding method for performing a motion search by a computing unit and coding a moving picture by using a motion compensation inter-frame predictive coding method, comprising: an image inputting step of inputting an image; A block motion search step for performing motion search for each area, an inter-frame prediction coding step for performing motion compensation inter-frame prediction coding using a motion vector for each area obtained in the motion search step, A local decoding step of decoding encoded data, comprising: setting a target encoding time which is a time for encoding one frame from a predetermined frame rate before encoding. A time setting step and a method for storing motion search instruction number information, which is the number of motion search processing instructions to be executed by the arithmetic unit, for a motion search method set in advance before encoding. A parameter storing step, a coding time measuring step for measuring the coding time of one frame, and a coding target time and coding time measuring step set in a target time setting step after the completion of one frame coding. From the measured coding time, a motion search time setting step for setting a motion search target time for performing a motion search for the next frame, and a motion search time for each region executed in the block motion search step are measured. A block motion search time measuring step, a block motion search time adding step for calculating a frame motion search time obtained by summing, for each frame, a motion search time for each block measured in the block motion search time measuring step, and a motion parameter storage step Target time and frame motion according to the number of motion search commands stored in Processing time adaptive image coding method characterized by executing a motion search control step of controlling the determined motion search motion estimation method from search time, the. 6. A moving picture coding method for performing a motion search and a global motion search in a computing unit and coding a moving picture using a motion compensated inter-frame predictive coding method, comprising: A block motion search step for dividing the screen into regions and performing a motion search for each region; a global block motion search step for dividing the screen into regions and performing a motion search for each region to obtain a global motion vector; An inter-frame predictive coding step of performing motion-compensated inter-frame predictive coding using the motion vector for each area obtained in the motion search step and the global motion vector obtained in the global block motion search step, and decoding of encoded data A video decoding method for performing a local decoding step of performing A target time setting step of setting an encoding target time which is a time for encoding one frame from a frame, and a motion search method set in advance before encoding.
A motion parameter storage step for storing information on the number of motion search instructions, which is the number of processing instructions for motion search executed by the arithmetic unit, and a motion executed by the arithmetic unit with respect to a global motion search method set before encoding. A global motion parameter storing step for storing information on the number of global motion search instructions, which is the number of search processing instructions, an encoding time measuring step for measuring the encoding time of one frame, and a target time after one frame encoding is completed From the coding target time set in the setting step and the coding time measured in the coding time measuring step, a local motion search target time, which is a motion search target time for performing a motion search of the next frame, and , A motion search time setting step for setting a global motion search target time which is a motion search target time for performing a global motion search The block motion search time measuring step for measuring the motion search time for each region executed in the block motion search step and the motion search time for each block measured in the block motion search time measuring step are summed up in frame units. A block motion search time adding step for obtaining a frame motion search time, a global block motion search time measuring step for measuring a motion search time for each area executed in the global block motion search step, and a global block motion search time measuring step. A local block motion search time adding step for obtaining a frame global motion search time obtained by summing the measured motion search times for each block, and a local motion search command number information stored in the motion parameter storage step. Motion search eyes A motion search control step for determining a motion search method from the time and the frame motion search time and controlling the motion search processing; and a global motion search target time according to the number of global motion search instructions stored in the global motion parameter storage step. A global motion search control step of determining a global motion search method from a frame global motion search time and controlling a global motion search process. 7. A moving picture coding method for performing a motion search by a computing unit and coding a moving picture by using a motion compensation inter-frame predictive coding method, comprising: a picture inputting step of inputting a picture; A block motion search step for performing motion search for each area, an inter-frame prediction coding step for performing motion compensation inter-frame prediction coding using a motion vector for each area obtained in the motion search step, A local decoding step of decoding encoded data, comprising: setting a target encoding time which is a time for encoding one frame from a predetermined frame rate before encoding. A time setting step and a method for storing motion search instruction number information, which is the number of motion search processing instructions to be executed by the arithmetic unit, for a motion search method set in advance before encoding. A block priority detecting step of detecting image statistical information of each area in the screen from image information of an input image, and a search block determining step of determining an order of a motion search area from the image statistical information. And an encoding time measuring step for measuring the encoding time of one frame, and after the encoding of one frame,
A motion search time for setting a motion search target time for performing a motion search for the next frame from the coding target time set in the target time setting step and the coding time measured in the coding time measuring step. The setting step, a block motion search time measuring step for measuring a motion search time for each area executed in the block motion search step, and a motion search time for each block measured in the block motion search time measuring step are calculated in frame units. A motion search method is determined from the target motion search time and the frame motion search time in accordance with the block motion search time addition step for obtaining the frame motion search time summed in the above and the motion search command number information stored in the motion parameter storage step. Performing a motion search control step of controlling the motion search process; Response image coding method. 8. A moving picture coding method for performing a motion search and a global motion search in an arithmetic unit and coding a moving picture using a motion compensated inter-frame predictive coding method, comprising the steps of: A block motion search step for dividing the screen into regions and performing a motion search for each region; a global block motion search step for dividing the screen into regions and performing a motion search for each region to obtain a global motion vector; An inter-frame predictive coding step of performing motion-compensated inter-frame predictive coding using the motion vector for each area obtained in the motion search step and the global motion vector obtained in the global block motion search step, and decoding of encoded data A video decoding method for performing a local decoding step of performing A target time setting step of setting an encoding target time which is a time for encoding one frame from a frame, and a motion search method set in advance before encoding.
A motion parameter storage step for storing information on the number of motion search instructions, which is the number of processing instructions for motion search executed by the arithmetic unit, and a motion executed by the arithmetic unit with respect to a global motion search method set before encoding. A global motion parameter storing step of storing information on the number of global motion search instructions, which is the number of search processing instructions, and a block priority detecting step of detecting image statistical information of each area in the screen from the image information of the input image. A search block determining step of determining an order of a region for motion search from image statistical information, an encoding time measuring step of measuring an encoding time of one frame, and a target time setting step after the end of one frame encoding. A motion search for the next frame is performed based on the set encoding target time and the encoding time measured in the encoding time measuring step. And a block motion search step for setting a local motion search target time, which is a motion search target time, and a global motion search target time, which is a motion search target time for performing a global motion search. Block motion search time measuring step for measuring the motion search time for each area, and the block motion for calculating the frame motion search time by summing the frame motion search time for each block measured in the block motion search time measurement step A global block motion search time measuring step for measuring a motion search time for each area performed in the search time adding step, a global block motion search step, and a motion search for each block measured in the global block motion search time measuring step The total time in frames The motion search method is determined from the local motion search target time and the frame motion search time according to the global block motion search time adding step for obtaining the global motion search time and the motion search command number information stored in the motion parameter storage step. A global motion search method is determined from a global motion search target time and a frame global motion search time according to the motion search control step for controlling the search process and the global motion search command number information stored in the global motion parameter storage step. A global motion search control step of controlling a motion search process. 9. Any one of claims 1 to 8
A recording medium for a processing time adaptive image encoding program, which stores a program for causing a computer to execute the processing time adaptive image encoding method described in the section.