JP2000287212A - Image encoder - Google Patents

Abstract

PROBLEM TO BE SOLVED: To eliminate changes in an image which is used to happen by a decode image difference of an I frame and to improve an image visual quality with respect to a still area part such as a background by reading a code of the I frame 1 GOP prior to a bit stream buffer in the case of a macroblock that is decided to be a still area and encoding the code as a code of the still area. SOLUTION: In a code amount control part 14, with respect to a macroblock which is decided to be a still area, a code amount of a corresponding macroblock in an I frame 1 GOP before is read from a code amount buffer 12 and integrated on the basis of a still/dynamic area decision result from an ME part 15. In an encoding part 17, encoding is performed on the basis of the still/dynamic area decision result and a target code amount supplied by the code amount control part 14 by macroblocks. In the case of the still area, a code of the corresponding macroblock in the I frame 1 GOP before in a bit stream buffer 13 is read and this is outputted as a code of a present macroblock.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

[0001] 1. Field of the Invention [0002] The present invention relates to image coding using MPEG.

[0002]

2. Description of the Related Art FIG. 3 shows a sequence of image coding using MPEG. Here, a plurality of frames constitute a unit called a GOP, of which one frame is subjected to intra-frame encoding called an I-frame, and the other frames are P-frames for performing forward prediction and B-frames for performing bidirectional prediction. Perform frame encoding. Each frame is divided into units called macroblocks. In the case of a P frame or a B frame, a motion vector is detected for each macroblock, a coding mode is determined, and a quantization value is determined. The prediction error, the motion vector, and the encoding mode are encoded. In the case of an I frame, a quantization value is determined for each macroblock, and coding is performed.

[0003] For a stationary part, in a P frame, a motion vector is set to zero and no prediction error is coded.
By selecting oded, in the decoded image, the image at the same position of the previous I frame is copied and output. In the B frame, the motion vector is set to zero,
By not sending the prediction error, it is possible to reproduce the same image as the previous I frame from the similarly decoded image. As described above, for the still portion, it is possible to encode all the frames in the GOP so as to obtain the same decoded image as the I frame.

However, since the next I frame is encoded in the frame, the encoding is performed irrespective of the I frame one GOP before. For this reason, regarding the stationary portion, the quantized value may be different due to code amount control, or the decoded image may be different even if the quantized value is the same due to noise or the like in the image. .
In such a case, the still portion becomes an image different for each GOP, and the change becomes visually conspicuous for each GOP, causing visual deterioration.

[0005]

With respect to the next I frame, encoding is performed irrespective of the I frame one GOP before. For this reason, the quantization value of the stationary portion is changed to a different value by code amount control. In some cases, even if the quantization value is the same due to noise in the image, a different decoded image may be obtained. In such a case, there is a problem that a still portion becomes a different image for each GOP, and therefore, the change becomes visually noticeable in GOP units, causing visual deterioration.

[0006]

In order to solve this problem, an image encoding apparatus for encoding an input image comprises: an I-frame before one GOP;
A frame buffer for accumulating the P I frame and the decoded reference frame, and when the current frame to be processed is an I frame, the determination of a still / moving region between the I frame before one GOP and the current I frame is performed for each macro block. A motion vector detecting unit, a code amount buffer for accumulating a generated code amount for each I-frame macroblock, and a bitstream buffer for accumulating a bitstream generated for each I-frame macroblock in macroblock units. , For the macroblock determined to be a still area by the motion vector detection unit, read out the generated code amount of the I frame one GOP before from the code amount buffer, calculate the sum thereof, and calculate the motion vector by the remaining code amount. A code for calculating a target code amount of a macroblock determined as a moving area by the detection unit. The amount control unit, the macroblock determined as the moving area is coded according to the target code amount, and in the case of the macroblock determined as the still area, the I-frame of one GOP before the G frame is 3. The coding apparatus according to claim 2, further comprising: a coding unit that reads a code and performs coding as a code of the still area; and a motion compensation unit that performs motion compensation prediction in the case of a P frame or a B frame. The I-frame of the GOP before 1 GOP, the I-frame of the current GOP
A frame buffer for storing a frame and a decoded reference frame, and a motion for determining a still / moving area for each macroblock between the I frame one GOP before and the current I frame when the current frame is an I frame. A vector detecting unit, a code amount buffer for storing a generated code amount for each macroblock of the I frame, a quantized value buffer for storing a quantized value for each macroblock of the I frame, For the macroblock determined to be an area, the generated code amount of the I frame one GOP before is read out from the code amount buffer, the sum is obtained, and the remaining code amount is determined to be a moving area by the motion vector detecting unit. A code amount control unit for calculating a target code amount of the macro block, and the macro block determined to be the moving region is Performs encoding in accordance with the target code amount,
In the case of the macroblock determined to be the still area, the quantization value of the corresponding macroblock of the I frame one GOP before the quantized value buffer and the corresponding macroblock of the I frame one GOP before the three frame buffer are determined. The encoding unit is configured to include an encoding unit that performs read encoding and a motion compensation unit that performs motion compensation prediction in the case of a P frame or a B frame.

[0007]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS One embodiment of the encoding apparatus according to the present invention will be described below with reference to FIG. As shown in FIG. 1, the encoding apparatus according to the present invention includes a frame buffer 11, a code amount buffer 12, a bit stream buffer 13, a code amount control unit 14, an ME unit 15, an MC unit 16, an encoding unit 17, and a local decoding unit. It comprises a unit 18.

The following processing is performed on the P frame and the B frame as in the conventional case. The input image is supplied to an ME unit 15 for detecting a motion vector and an MC unit 16 for performing motion compensation prediction.

In the ME section 15, a local decoding section 18
Then, block matching is performed in units of macroblocks between the reference image locally decoded by the reference image and stored in the three-frame buffer 11 and the input image to detect an optimal motion vector. The detected motion vector is the MC unit 1
6. The MC unit 16 determines a prediction mode for each macroblock from the supplied motion vector, the input image, and the reference image in the three-frame buffer 11, and determines the prediction mode, the motion vector, and the prediction error in the encoding unit 1.
7

The encoding unit 17 quantizes the prediction error based on the prediction mode, the motion vector and the prediction error supplied from the MC unit 16 and the quantization value specified by the code amount control unit 14, The encoding is performed together with the motion vector to generate and output a bit stream.

In the case of an I frame, the input image is supplied to the three-frame buffer 11 and the ME unit 15, and the input image is stored in the three-frame buffer 11 before encoding. The ME unit 15 is supplied with the input image and the previous I frame previously stored in the three-frame buffer 11, and detects a motion vector for each macroblock. If the detected motion vector is zero, the code amount control unit 14 determines that the region is a still region, and if the detected motion vector is not zero, the region is a motion region.
Is supplied with information on the determination result of the stationary / moving region.

The code amount control unit 14 determines the code amount of the corresponding macroblock in the I frame one GOP before the macroblock determined as the still region based on the still / moving region determination result from the ME unit 15. The data is read from the code amount buffer 12 and integrated. A value obtained by subtracting the current target code amount (Ia) of the current I frame from the integrated code amount (Is) for all the I-frame macroblocks is integrated (Im) of the target code amount of the macroblock determined as the moving area. And the encoding unit 1
7 is subjected to code amount control.

The coding unit 17 performs coding based on the still / moving area determination result and the target code amount supplied from the code amount control unit 14 for each macroblock. In the case of the still area, the code of the corresponding macroblock of the I frame one GOP before in the bit stream buffer 13 is read, and this is output as the code of the current macroblock.

In the case of the moving area, the macro block of the I frame of the current GOP is read from the three-frame buffer 11 and is encoded so as to have the target code amount. The code is output as a bit stream for each generated macro block, and is stored in the bit stream buffer 13 for each macro block.

The generated code amount is stored in the code amount buffer 12 for each macroblock. Also, if the accumulated code amount (Is) is large relative to the current target code amount (Ia) of the I frame, the virtual buffer underflows, so that encoding cannot be performed, so that all macroblocks are activated. All macro blocks are coded as a region.

Another embodiment of the image coding apparatus according to the present invention will be described below with reference to FIG. As shown in FIG. 2, the encoding device according to the present invention comprises a frame buffer 11,
Code amount buffer 12, code amount control unit 14, ME unit 15,
It comprises an MC unit 16, an encoding unit 17, a local decoding unit 18, and a quantization value buffer 19. In FIG. 1, the code for each macroblock is stored, but in FIG. 2, the quantization value for each macroblock is stored.

Since the quantization value can be represented by 5 bits of values from 0 to 31, the amount of memory can be reduced. In this case, for the still region, the encoding unit 17 reads the corresponding macroblock from the I frame one GOP before in the three-frame buffer 11 and the quantization value from the quantization value buffer 19, and performs encoding. For all macroblocks, the quantization value is stored in the quantization value buffer 1
9 for each macroblock.

[0018]

According to the present invention, the change in the image caused by the difference between the decoded image of the I frame and the still area such as the background is eliminated, and the visual quality is improved.

[Brief description of the drawings]

FIG. 1 is a configuration diagram showing an embodiment of an encoding device of the present invention.

FIG. 2 is a configuration diagram showing another embodiment of the encoding device of the present invention.

FIG. 3 is a diagram showing a sequence of image encoding using MPEG.

[Explanation of symbols]

 Reference Signs List 11 frame buffer 12 code amount buffer 13 bit stream buffer 14 code amount control unit 15 ME unit 16 MC unit 17 encoding unit 18 local decoding unit 19 quantized value buffer

Claims (2)

[Claims] 1. An image coding apparatus for coding an input image, comprising: an I frame before one GOP, an I frame of a current GOP,
And a frame buffer for storing the decoded reference frame, and 1 GO if the current frame to be processed is an I frame.
A motion vector detecting unit that determines a still / moving region between the I frame before P and the current I frame for each macroblock, a code amount buffer for accumulating the generated code amount for each macroblock of the I frame, A bit stream buffer that accumulates a bit stream generated for each macro block of the frame in macro block units, and a generated code amount of an I frame one GOP before the macro block determined as a still area by the motion vector detection unit. Is read from the code amount buffer, the sum thereof is obtained, and, with the remaining code amount, a code amount control unit that calculates a target code amount of a macroblock determined to be a moving region by the motion vector detection unit; The determined macroblock is coded according to the target code amount, and the macroblock determined to be the stationary area is encoded. In the case of a block, a coding unit that reads the code of the I frame one GOP before from the bit stream buffer and performs coding as the code of the still region, and a motion compensation unit that performs motion compensation prediction in the case of a P frame or a B frame. An image encoding device comprising: 2. An image encoding apparatus for encoding an input image, comprising: an I-frame before one GOP, an I-frame of a current GOP,
A frame buffer for storing the decoded reference frame; and a motion vector detecting unit for determining a still / moving area for each macroblock between the I frame before one GOP and the current I frame when the processing frame is an I frame; A code amount buffer for storing a generated code amount for each macroblock of the I frame; a quantized value buffer for storing a quantized value for each macroblock of the I frame; For the macroblock, the generated code amount of the I frame one GOP before is read from the code amount buffer, and the sum thereof is obtained.
A code amount control unit that calculates a target code amount of a macroblock determined to be a moving region by the motion vector detection unit with the remaining code amount; and a macroblock determined to be a moving region corresponds to the target code amount. In the case of a macroblock determined to be the still area, 1 G
A coding unit for reading out and coding the quantized value of the corresponding macroblock of the I-frame before the OP and the corresponding macroblock of the I-frame before the GOP from the three-frame buffer; An image coding apparatus comprising: a motion compensation unit that performs motion compensation prediction.