JP2000032474A - Dynamic image encoder - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a dynamic image encoder capable of making a GOP configuraltion adaptively variable, while considering the effects of motion vector search, and preventing encoding operation from being delayed so much, without needing any separate memory and computing element. SOLUTION: Input images are stored in an encoding image memory 4 and successively outputted to a motion vector detection circuit 6. In the motion vector detection circuit 6, separate from prediction for encoding, a motion vector between adjacent frames is also detected for plural blocks. In an encoding type selection circuit 7, an appropriate encoding type is selected from intra- frame encoding, inter-frame encoding in a forward direction and inter-frame encoding in both directions from the inputted motion vector between the adjacent frames. A memory controller 5 controls the input/output of the encoding image memory 4, corresponding to the selected result of the encoding type selection circuit 7, and an encoding part 2 encodes the input images in accordance with the decision result of the encoding type selection circuit.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to encoding of a moving image using intra-frame encoding, forward inter-frame encoding, and bi-directional inter-frame encoding. The present invention relates to a moving picture coding apparatus using a motion compensation inter-frame coding method performed by a motion vector search circuit that performs a motion vector search.

[0002]

2. Description of the Related Art In recent years, MPE has been used as a moving picture coding method.
G1 (ISO / IEC 11172), MPEG2 (I
An interframe coding method using motion compensation prediction such as SO / IEC 13818) is used in the fields of storage, communication, and broadcasting. Although the field structure is defined in MPEG2, the description will be limited to the frame structure here. In the MPEG system, intra-frame coding, forward inter-frame coding, and bi-directional inter-frame coding are switched and used for each frame. Hereinafter, a frame to be intra-coded is referred to as an I frame, a frame to be subjected to inter-frame coding in the forward direction is referred to as a P frame, and a frame to be subjected to inter-frame coding in both directions is referred to as a B frame.

FIG. 17 shows an example of the arrangement of an I frame, a P frame, and a B frame. In the MPEG system, as shown in FIG. 17, a plurality of frame groups including one or more I frames are grouped into Group Of Pictures (hereinafter referred to as GO).
P). In a GOP, an I frame and a P frame are generally arranged at equal intervals, and the interval between them is a B frame. Also, an image is called a macroblock.
The image is divided into blocks of pixels × 16 pixels, and prediction between frames is performed for each macroblock.

In general, the structure of a GOP is often fixed, but by making it variable, coding more suitable for the characteristics of the image can be performed. For example, in a continuous scene or a scene with a small motion, high image quality can be obtained with a small code amount by performing inter-frame coding, but a case where the correlation between frames becomes extremely low due to a scene change or a scene with a large motion. In this case, the coding efficiency is improved by using the intra-frame coding rather than the inter-frame coding.

Japanese Patent Application Laid-Open No. 8-65678 discloses a system in which the number of frames constituting a GOP and the interval between I and P frames are made variable in accordance with the characteristics of an image. In this patent, an approximate expression of the reciprocal of the coding efficiency (Gain) is derived,
A GOP configuration that minimizes the value of the approximate expression is required. Expression (1) shows an approximate expression.

[0006]

(Equation 1)

In equation (1), ρ is a correlation coefficient between adjacent frames, WP and WB are weighting factors determined by a ratio of a quantization step width set for each frame, and m is I,
The interval between P frames, n is the number of P frames, N is the maximum number of GOP frames that can be set, and S (m-1) is an expression represented by the following expression (2).

[0008]

(Equation 2)

In the technique described in Japanese Patent Application Laid-Open No. 8-65678, an input image is temporarily stored in a memory, and after searching for a combination of m and n that minimizes the value of equation (1),
By using a GOP configuration according to the obtained values of m and n, a GOP with high encoding efficiency is obtained.

[0010]

However, in the method described in JP-A-8-65678, the GOP configuration is determined based on the correlation between frames, and the effect of motion prediction is not considered. When using inter-frame prediction, the image quality is greatly affected by the result of motion prediction. For example, when the motion amount is large and is out of the range of the motion vector search, there is a problem that the effect of inter-frame prediction is not exhibited and the image quality is extremely deteriorated.

In the case of the GOP structure shown in FIG. 17, the prediction of a P frame is a prediction from a frame three frames away. As the interframe distance increases, the motion amount also increases, and the motion amount often falls outside the motion vector search range. In such a case, the image quality of the P frame becomes extremely poor, and further prediction is performed from an image with poor image quality. Therefore, the next prediction also enters a vicious cycle in which the prediction error increases. Therefore, it is desirable to adopt a GOP configuration in which when the motion amount is large, the prediction between the frames is small.

In the method described in JP-A-8-65678, it is necessary to temporarily store an input image in a memory in GOP units. The GOP is generally 12 frames or 15 frames. For example, the image format is 4: 2:
2. When the image size is 720 x 480 pixels, the data amount of one screen is 720 x 480 x 2 = 691,200 (approximately 700K Byte). There is a problem that a large capacity memory is required.

Further, since the data is temporarily stored in the memory, a delay occurs from the input of data to the start of encoding. 1G
When the OP is 15 frames, the delay is about 0.5 seconds in the case of the NTSC signal. For applications requiring real-time properties such as broadcasting and communication, delay is a problem.

Further, in order to determine the GOP configuration, it is necessary to repeatedly calculate a correlation coefficient between frames or to perform calculations including division and factorial, and thus the amount of calculation is extremely large.

The present invention has been made in view of such a problem, and has been developed in consideration of the effect of motion vector search.
It is an object of the present invention to provide a moving picture coding apparatus which makes an OP configuration adaptively variable, does not require a separate memory or arithmetic unit, and does not increase coding delay.

[0016]

According to the present invention, a correlation operation is performed between adjacent frames, and a GOP structure suitable for an image sequence is selected using the result of the correlation operation. In determining the GOP configuration, the effect of motion prediction is also taken into consideration. If the motion amount is large and prediction with a large inter-frame distance is inappropriate, the GOP configuration is such that prediction is made to shorten the inter-frame distance. .

Further, the correlation calculation is performed only on a part of the screen to reduce the calculation amount of the correlation calculation, and furthermore, the correlation calculation is sequentially performed on the input frame by using the existing motion vector detecting device, so that the delay amount is reduced. , And the GOP configuration can be made variable without increasing the circuit scale for the correlation operation.

According to the first aspect of the present invention, there is provided a moving picture coding apparatus for coding by combining intra-frame coding, forward inter-frame coding, and bi-directional inter-frame coding. A coded image memory for storing a coded frame and a frame adjacent to the frame, and dividing the coded frame into blocks, selecting a plurality of blocks, and for each selected block, a frame adjacent to the frame. A motion vector detecting unit that performs an error operation and detects a first motion vector; and a coding type that selects a coding type of a coded frame based on the first motion vector detected by the motion vector detecting unit. A selecting unit, a coded frame read from the coded image memory based on the selected coding type, and a coded frame read from the reference image memory. Second motion vector by performing an error calculation for each block of the reference image to detect the issue, the second
The above problem is solved by providing an encoding unit that generates encoded data from the motion vector and the read reference image.

According to a second aspect of the present invention, the encoding section detects a second motion vector in the motion vector detecting section based on the encoding type selected by the encoding type selecting section. The above problem is solved by generating encoded data from the detected motion vector and the reference image.

According to the third aspect of the present invention, when the search range of the first motion vector is ± N pixels in the vertical and horizontal directions, the vertical and horizontal When the number of motion vectors included in the area of ± N / M pixels in the direction is equal to or greater than a predetermined number, the encoding type selection unit changes the encoding type of the M-1 frame including the encoded frame between the bidirectional frames. The above problem is solved by encoding.

According to the fourth aspect of the present invention, when the search range of the first motion vector is ± N × ± 2N pixels, (± N / M) ×
When the number of motion vectors included in the area of (± 2N / M) pixels is equal to or less than a predetermined number, the encoding type selection unit sets the encoding type of the M-1 frame including the encoded frame to the forward frame. The above problem is solved by performing inter-coding and detecting the second motion vector as a search range of ± N × ± 2N pixels.

According to the fifth aspect of the present invention, the motion vector detecting unit performs an error calculation for each selected block with respect to a frame adjacent to the selected frame, and calculates an error corresponding to the first motion vector and the vector. The above problem is solved by calculating an error value and, when the error value is equal to or more than a predetermined value, invalidating the obtained first motion vector and not using the first motion vector for determination in the coding type selection unit.

According to the sixth aspect of the present invention, the motion vector detecting unit performs an error operation between the selected block word and a frame adjacent to the frame, and calculates an error corresponding to the first motion vector and the vector. Find the error value
When the number of motion vectors whose error value is equal to or greater than a predetermined value is equal to or greater than a predetermined number, the above-mentioned problem is solved by selecting intra-frame encoding as the encoding type of the frame in the encoding type selection unit. I do.

According to the seventh aspect of the present invention, at least a position of a frame to be intra-coded in the plurality of frames is set in advance for each of a plurality of frames. The above problem is solved by selecting an encoding type in the encoding type selection unit.

[0025]

Embodiments of the present invention will be described below in detail with reference to the drawings. FIG. 1 is a block diagram showing a first embodiment of the present invention. This is an embodiment corresponding to claim 1. The conventional video coding device is GO
Although an encoding type for each frame in P is fixed or variable, an arithmetic unit for performing a correlation operation and a large-capacity memory are required. In the present embodiment, encoding is performed according to image characteristics. It is characterized in that, although the type of conversion is variable, no additional correlation calculator or memory is required.

In this embodiment, the MPE shown in the conventional example
An example in which encoding is performed by the G method and the number of B frames between the I and P frames is set to at most one will be described.

In the present embodiment shown in FIG. 1, a coding type determining unit 1 for adaptively selecting a coding type for each frame from intra-frame coding, forward inter-frame coding, and bidirectional inter-frame coding. An encoding unit 2 for encoding an image, and a local decoding unit 3 for generating a decoded image used for inter-frame prediction or predicted image generation. This embodiment is particularly characterized by the coding type determination unit 1.

In FIG. 1, an input image is input to a coded image memory 4. The encoded image memory 4 has a capacity of three frames, and the input and output are controlled by the memory controller 5. The operations of the memory controller 5 and the encoding type selection circuit 7 will be described later. A frame to be coded is read from the coded image memory 4, a reference image is read from the reference image memory 14, and a motion vector is detected by the motion vector detection circuit 6 for each macroblock of 16 × 16 pixels. You.

After the detection of the motion vector, the pixel data of the macroblock to be coded from the motion vector detection circuit 6 is
From the reference image memory 14, a predicted image obtained from the encoded macroblock position and the motion vector is input to the subtracter 8, and a prediction error image is input to the DCT circuit 9.

The DCT circuit 9 converts the prediction error data into weight coefficients in the frequency domain, the quantizer 10 quantizes the weight coefficients, and the quantized result is subjected to variable-length coding by the variable-length coding circuit 11, It is output as coded data. The quantization result is also input to the inverse quantization circuit 12, and the inverse quantization result is converted into prediction error data by the inverse DCT circuit 13.
The prediction error data is added to the prediction image by the adder 15 to form a decoded image. The decoded image is stored in the reference image memory 14 and is used as a reference image or a predicted image.

An encoding type determining unit 1 which is a feature of the present invention.
Will be described in more detail. First, FIGS. 2A to 2C show the writing and reading operations of the encoded image memory 4. 1
The frame period is divided into an encoding period and a determination period, and an input frame is divided into a frame F1 and a period T2 during the encoding period of the period T1.
Are input in the order of frame F2. During the encoding period, writing and reading of the input frame are performed, and during the determination period, the encoding type of the immediately preceding input frame is determined. In FIG. 2C, the subscript “W” indicates that writing to the memory is performed, and “R” indicates that reading from the memory is performed.

The writing and reading of the coded image memory 4 are controlled by the memory controller 5 in accordance with the flowcharts shown in FIGS. FIG. 3 shows a flowchart of the write operation, and FIG. 4 shows a flowchart of the read operation.

In the flow chart of the write operation shown in FIG.
To set the write address of the encoded image memory 4 to M1, M
2, and M3, and sequentially write input frames for three frames. Although not shown in the flowchart of FIG. 3, both writing and reading are performed from the third frame. Then, the memory address which has been read immediately before and is vacant is updated as the next write address (step 1).
04), This operation is repeated until the input frame ends. That is, in the third and subsequent frames, the next frame is written to the memory address from which reading has been completed.

Next, the read operation will be described. The reading is different between the I and P frames and the B frame. The I and P frames may be read in the order of writing, but the B frame needs to be read after the I and P frames following the B frame.

Therefore, the FI for storing the write address is used.
When the coding type is determined to be I or P frame, the write address is directly input to the FIFO. When the coding type is determined to be B frame, the write address of the I and P frames following the B frame is set to the FIFO. After the writing, the write address of the B frame is input to the FIFO, and reading is performed in the order of output from the FIFO.
This operation is shown in the flowchart of FIG. 4, and examples of the state of the FIFO are shown in FIGS. 2 (g), (h) and (i).

In FIG. 4, the encoding type of the frame written immediately before at the address Mi is selected (step 111). The encoding type selection means will be described later. If the selection result is a B frame, the address Mi is set to FIFO.
It waits for input (step 113). If the selection result is not a B frame, that is, if it is an I or P frame, it is determined whether there is a FIFO input wait (step 114). If there is no input wait, Mi is input to the FIFO. (Step 11
6) If there is an input waiting, input to the FIFO in the order of Mi, FIFO input waiting address (step 11).
5). The FIFO output is used as a read address (step 117), and reading is performed in the next encoding period.

FIG. 4 is a flowchart for determining the read address of one frame. This operation is repeated until there are no more read frames. In step 113, the address where the B frame is written is waiting for FIFO input. In step 115, the address waiting for FIFO input is the FIFO next to the address where the I and P frames are written.
, The B frame is read after the subsequent I and P frames.

Next, the operation of the encoding type selection circuit 7 will be described. The coding type selection circuit 7 selects the coding type of the frame input during the coding period in the determination period following the coding period (FIGS. 2C and 2D). The encoding type is selected from the encoding type candidates shown in FIG. The selection of the encoding type is performed using the result of correlation calculation between adjacent frames and the result of motion vector detection input from the motion vector detection circuit 6.

First, FIG. 5 is a flowchart showing the operation of determining a coding type candidate. Next, means for selecting a coding type from the coding type candidates will be described.

In FIG. 5, first, the coding type candidate of the first frame is set to I or B (step 131). This is because the first frame has no prediction image and cannot perform forward prediction. Next, it is determined whether or not there is an input frame next to the frame for determining the coding type candidate (step 132).

If there is an input frame, it is determined whether or not the frame preceding the frame for determining the candidate is a B frame (step 133). In the case of a B frame, the coding type candidate is set to I or P (step 134). This is because two or more B frames are not continuous in the present embodiment. If the previous frame is not a B frame, the coding type candidate is set to I, P or B, that is, all types are set as candidates (step 135).

If there is no next frame in step 132, the coding type candidate is set to I or P (step 136). This is because the last frame cannot be a B frame predicted from the rear.

Next, the concept of selecting an encoding type from encoding type candidates will be described with reference to FIGS. FIG. 6 shows an example of the motion vector search range. FIG. 6 shows a case where the motion vector search range is −M to + M pixels in both the horizontal and vertical directions. FIG. 6A shows an example of the search range of the B frame.
(B) shows an example of the search range of the P frame example. In this example, assuming that the value of the horizontal motion vector between the frames 1 and 2 is (M / 2 + 1), the value of the horizontal motion vector between the frames 1 and 3 is (M + 1). Can be expected.

However, the horizontal search range between the frames 1 and 3 is -M to + M, and the motion vector is outside the search range. As a result, it is expected that a non-optimal motion vector will be selected and the prediction error will increase.

In such a case, by changing the frame 2 from the B frame to the P frame as shown in FIG. 7, the horizontal search range between the adjacent frames is -M to + M, and the frame search between the frames 1 and 3 is the frame. 2, a search in the range of -2M to + 2M becomes possible, and a motion vector outside the range can be searched accurately.

Therefore, a motion vector having an inter-frame distance of 2 is predicted from a plurality of motion vectors between adjacent frames,
Even when the inter-frame distance is 2, many vectors are predicted to be within the search range. If the encoding type candidate includes a B frame, the B frame is selected. If there are many vectors outside the search range, the vector is selected. An encoding type is selected so that I and P frames are continuous. For example, in the case of FIG. 6B, the result of the motion vector search between the frames 1 and 3 is predicted from the motion vector between the frames 1 and 2, and the coding type of the frame 2 is selected from the coding type candidates. Become.

To determine whether to select a B frame,
The motion vector search range between adjacent frames is divided into regions,
This is performed using the number of motion vectors included in each area. FIG. 8 shows an example in which a motion vector search range between adjacent frames is divided into regions. FIG. 8 shows an example in which the image is divided into two regions. However, the image may be divided into more regions. In the case where the vector between adjacent frames is included in the first region in FIG. 8, the motion vector of the P frame is within the search range even when the distance between the frames is 2, that is, −M to + M pixels in both the horizontal and vertical directions. It can be predicted that the motion vector is within the range, and if the vector between adjacent frames is included in the second area, it can be predicted that the motion vector whose inter-frame distance is 2 is outside the search range.

The coding type selection circuit 7 performs motion vector detection between adjacent frames for some macroblocks in a frame such as 9 macroblocks or 16 macroblocks during the determination period shown in FIG. To select the encoding type. FIGS. 9A and 9B show examples of selecting macroblocks when error calculation is performed on 9 macroblocks and 16 macroblocks, respectively. The hatched portion is a macroblock for which error calculation is performed.

As shown in FIG. 2, in the present embodiment, the determination period is set to a period that does not overlap with the encoding period. Accordingly, by controlling the motion vector detection circuit 6 to perform a motion vector detection for encoding during the encoding period and to perform an error operation for selecting the encoding type during the determination period, the coding type selection is performed. It is not necessary to newly add a motion vector detection circuit for the above.

FIG. 10 is a flowchart showing the operation of selecting an encoding type. The encoding type selection circuit 7 sets a plurality of macroblocks in the horizontal direction between adjacent frames.
A motion vector searched for a range of -M to + M pixels in the vertical direction is input, and an optimal coding type is selected. The motion vector input to the encoding type selection circuit 7 is the motion vector between the input frame and the frame immediately before the input frame. For example, during the determination period of T4 in FIG. 2, the motion vector of the frame F4 using the frame F3 as the reference image Is input, and the encoding type of the frame F4 is selected.

In FIG. 10, it is determined whether or not a coding type candidate includes a B frame (step 1).
41). If it is included, the number of motion vectors included in the first area shown in FIG. 8 is counted, and if the value is equal to or greater than a preset first threshold, B frame encoding is selected ( Step 143). This is because when the inter-frame distance is 1, if there are many motion vectors included in the first area, it can be expected that there are many motion vectors included in the motion vector search range even when the inter-frame distance is set to 2. is there. If the value is equal to or smaller than the first threshold value in step 142, P-frame encoding for motion vector search between I frames or adjacent frames is selected. In step 142, a second threshold value may be set in advance to determine whether the number of motion vectors included in the second area is equal to or less than the second threshold value. The selection between the I frame and the P frame is made by using the reliability of the motion vector as described later, or using a method of setting the I frame at regular intervals as shown in the second embodiment. be able to.

Next, the concept of selecting the coding type using the reliability of the motion vector will be described. The reliability of the motion vector is determined, for example, by inputting an evaluation value of the motion vector selection together with the motion vector from the motion vector detection circuit 6 to the coding type selection circuit 7 and determining the evaluation value. For example, when the sum of absolute differences of pixels between blocks is used as the evaluation value, a function is prepared that determines that the smaller the absolute difference is, the higher the reliability of the motion vector is. Then, a third threshold value is set in advance, and a motion vector whose reliability is equal to or less than the third threshold value is set as an error vector. If there are many error vectors, it is determined that the motion vector search itself is unreliable. Frame coding is used.

FIG. 11 is a flowchart showing the operation of selecting an encoding type using the reliability of a motion vector. This operation is performed after step 144 in FIG. Also, B
The determination may also be made after the frame is selected (step 143), and when there are many error vectors, it may be determined that the frame is an I frame. Referring to FIG.
The number of error vectors is counted at 1, and if the error vector is equal to or larger than a preset fourth threshold value, it is considered that inter-frame prediction has not been performed, so that I-frame encoding is performed without using prediction (step 152). When the number of error vectors is less than the fourth threshold, it is considered that the prediction error amount is small and the prediction is effective. Therefore, the coding type is determined (step 153), and in the case of the B frame, the B frame is coded as it is (step 154), and in the case of the I or P frame, the P frame is coded (step 155).

Next, a second embodiment of the present invention will be described. As shown in FIG. 6, the search range of the motion vector in the first embodiment is −M to + M pixels in both the horizontal direction and the vertical direction in both the P and B frames. In the present embodiment, the search range is illustrated in FIG. Thus, the feature is that the horizontal search range of the P frame is set to −2M to + 2M pixels.

This is because the B frame needs to search for -M to + M pixels from the forward direction and the backward direction, respectively, and the motion vector detection circuit sets the range of -M to + M pixels to two.
It has the processing ability to search multiple times. In the present embodiment, even in the P frame, all of this processing capability is utilized, and one of the search ranges in the horizontal direction and the vertical direction is doubled.
2M to + 2M pixels.

FIG. 13 shows an example of area division when the horizontal motion vector search range is extended to −2M to + 2M pixels. FIG. 14 shows a vertical motion vector search range of −2M pixels.
An example of area division in a case where the area is expanded to + 2M pixels will be described. In this embodiment and the first embodiment, as shown in FIGS. 13 and 14, the region division used for the coding type determination is different, but the other operations are the same.

Next, a third embodiment of the present invention will be described. In the first embodiment, the intervals between I frames are not uniform, but depending on the application, it is better to insert I frames at equal intervals and at high frequency. For example, when recording on a disk medium and playing back from the middle by random access, it is necessary to decode from an I frame, and when dividing encoded data by editing, it is necessary to make the I frame a break between divisions. In consideration of random access, it is desirable that the I frames be equally spaced so that it is easy to predict which frame is the I frame, and in consideration of editing, many I frames are inserted so that they can be edited in short units. It is desirable.

Therefore, in the third embodiment, I frames are made to appear at equal intervals, and more I frames are made to appear depending on the image sequence. This can be realized by setting a fixed GOP configuration in advance, changing the coding type from B and P frames to I, P and B frames, and not changing the coding type of I frames.

The third embodiment will be described with reference to FIGS. FIG. 15 is a flowchart showing an operation of determining a coding type candidate. The flowchart of FIG. 15 is different from the flowchart of FIG. 5 corresponding to the first embodiment in that in FIG. 15, it is determined whether or not the encoding type set in advance in step 163 is an I frame. The only difference is that a type candidate is also processed with only an I frame. With this process,
If a fixed GOP is set in advance, I frames will appear at equal intervals.

Next, the operation of selecting an encoded frame will be described with reference to FIG. The flowchart of FIG. 16 is different from the flowchart of FIG. 10 corresponding to the first embodiment in that it is determined in step 171 whether or not the encoding type candidate is only an I frame. The only difference is that a process for converting the data is added. By this processing, I-frame encoding is performed at equal intervals.

In the third embodiment, the control of the coded image memory 4 may be performed according to the flowcharts shown in FIGS. 3 and 4 in the first embodiment, and no special processing is required.

The specific example shown in the present embodiment is an example. In addition to the specific example, if the coded image memory has a larger capacity, a GOP in which two or more B frames
A configuration may be adopted. In that case, the calculation formula of the approximate value of the motion vector and the area division in FIG. 8 may be changed correspondingly. Further, the macroblock for searching for a motion vector shown in FIG. 9 is also an example, and a macroblock other than the one shown in FIG. 9 may be used. It can be a sum of squares.
Further, the present invention is not limited to the encoding system of the MPEG system, but can be applied to all motion compensation inter-frame encoding systems.

[0063]

According to the first aspect of the present invention, it is possible to add a large-capacity memory and a complicated arithmetic circuit to a conventional moving picture coding apparatus only by adding a coding type selection circuit. Encoding can be performed with an encoding type that matches the characteristics. In addition, since the present invention sequentially determines the coding type of an input image, the coding delay does not increase as compared with a conventional moving picture coding apparatus in which the coding type is fixed. Therefore, it can be used for applications other than the storage system, such as communication and broadcasting. Also, since the position of the I frame can be inserted not only at one fixed position but also at an arbitrary position, for example, even in a sequence where scene changes frequently occur, the I frame is inserted in response to the scene change. be able to.

According to the second aspect of the present invention, since the motion vector to be determined for determining the coding type and the motion vector to be used for encoding are obtained with the same configuration, the configuration can be simplified.

According to the third aspect of the present invention, it is possible to predict a motion vector detection result between frames separated by two or more frames from a motion vector between adjacent frames, and to select an encoding type adaptively for any GOP structure. Effect can be exhibited.

According to the fourth aspect of the present invention, since the motion vector between adjacent frames is detected by using all of the motion vector detection circuits used for bidirectional prediction, the motion vector can be searched over a wider range and the accuracy can be improved. The encoding type can be selected.

According to the fifth and sixth aspects of the present invention, when the motion vector is unreliable, a more appropriate coding type can be selected because the I-frame does not use the motion vector search.

According to the seventh aspect of the present invention, since the I-frame encoding appears every fixed frame even after the encoding type is changed, it is easy to divide the encoded data by reproduction or editing by random access. be able to.

[Brief description of the drawings]

FIG. 1 is a block diagram of a first embodiment of the present invention.

FIG. 2 is a diagram illustrating an operation of an encoded image memory.

FIG. 3 is a flowchart illustrating a write operation of an encoded image memory.

FIG. 4 is a flowchart showing a read operation of an encoded image memory.

FIG. 5 is a flowchart illustrating an operation of determining a coding type candidate.

FIG. 6 is a diagram illustrating an example of a motion vector search range.

FIG. 7 is a diagram illustrating an example of a motion vector search range when a GOP configuration is changed.

FIG. 8 is a diagram illustrating an example in which a motion vector search range is divided into regions.

FIG. 9 is a diagram showing a macroblock for which a motion vector search is performed to select an encoding type.

FIG. 10 is a flowchart showing an operation of selecting an encoding type.

FIG. 11 is another flowchart showing an operation of selecting an encoding type.

FIG. 12 is a diagram illustrating a motion vector search range according to the second embodiment of the present invention.

FIG. 13 is a diagram illustrating an example in which the motion vector search range is divided into regions when the motion vector search range is extended in the horizontal direction.

FIG. 14 is a diagram illustrating an example in which the motion vector search range is divided into regions when the motion vector search range is extended in the vertical direction.

FIG. 15 is a flowchart illustrating an operation of determining a coding type candidate according to the third embodiment of the present invention.

FIG. 16 is a flowchart illustrating an operation of selecting an encoding type according to the third embodiment of the present invention.

FIG. 17 is a diagram illustrating an example of a GOP configuration.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 Encoding type determination part 2 Encoding part 3 Local decoding part 4 Encoded image memory 5 Memory controller 6 Motion vector detection circuit 7 Encoding type selection circuit 8 Subtractor 9 DCT circuit 10 Quantization circuit 11 Variable length encoding circuit 12 Inverse quantization circuit 13 Inverse DCT circuit 14 Reference image memory 15 Adder

Continued on the front page F-term (reference) 5C057 CC04 EB18 ED07 EF05 EG06 EG08 EH07 EL01 EM00 EM04 EM09 EM13 EM16 GG01 GG06 GG07 5C059 MA00 MA05 MA14 MA23 MC15 NN01 NN03 NN28 PP05 PP06 PP07 SS11 TC03 TC03 TB03 TC03

Claims (7)

[Claims] 1. A moving picture coding apparatus for coding by combining intra-frame coding, forward inter-frame coding, and bi-directional inter-frame coding, comprising: inputting an input image; An encoded image memory for storing an adjacent frame; dividing the encoded frame into blocks, selecting a plurality of blocks, performing an error operation for each selected block with a frame adjacent to the frame, A motion vector detecting unit that detects a motion vector; a coding type selecting unit that selects a coding type of a coded frame based on the first motion vector detected by the motion vector detecting unit; Each frame of an encoded frame read from the encoded image memory and a reference image read from the reference image memory, based on the encoding type. And an encoding unit that performs an error operation for each of the blocks to detect a second motion vector, and generates encoded data from the second motion vector and the read reference image. Video encoding device. 2. The encoder according to claim 1, wherein the encoder detects a second motion vector in the motion vector detector based on the encoding type selected by the encoding type selector. 2. The moving picture encoding apparatus according to claim 1, wherein encoded data is generated from the reference picture. 3. When the search range of the first motion vector is ± N pixels in the vertical and horizontal directions, of the first motion vector, ± N / M pixels in the vertical and horizontal directions When the number of motion vectors included in the region is equal to or greater than a predetermined number, the encoding type selection unit sets the encoding type of the M-1 frame including the encoded frame to bidirectional inter-frame encoding. 3. The method according to claim 1, wherein
The moving picture encoding device according to the above. 4. The search range of the first motion vector is ±
When N × ± 2N pixels, among the first motion vectors, (± N / M) × (± 2
When the number of motion vectors included in the area of (N / M) pixels is equal to or less than a predetermined number, the coding type selecting unit sets the coding type of the M-1 frame including the coded frame to the forward inter-frame code. 3. The method according to claim 1, wherein the second motion vector is detected as a search range of ± N × ± 2N pixels.
The moving picture encoding device according to the above. 5. The motion vector detection unit performs an error operation with respect to a frame adjacent to the frame for each selected block to obtain the first motion vector and an error value corresponding to the vector. Is greater than or equal to a predetermined value, the obtained first motion vector is invalidated,
5. The moving picture coding apparatus according to claim 1, wherein the moving picture coding apparatus is not used for determination in the coding type selection unit. 6. The motion vector detection unit calculates an error between the selected block word and a frame adjacent to the frame to obtain the first motion vector and an error value corresponding to the vector. The method according to claim 1, wherein, when the number of motion vectors whose value is equal to or greater than a predetermined value is equal to or greater than a predetermined number, the encoding type selection unit selects intra-frame encoding as the encoding type of the frame. 5. The moving picture coding apparatus according to any one of claims 1 to 4. 7. For each of a plurality of frames, at least a position of a frame to be intra-coded in the plurality of frames is set in advance, and a code other than the set frame to be intra-coded is set in the coding type selection unit. 7. The moving picture coding apparatus according to claim 1, wherein a coding type is selected.