JP2010135885A - Image coding apparatus and method - Google Patents

Abstract

MPEG and H.264 In image coding such as H.264, the memory transfer amount is large, and the circuit scale and power consumption are very large.
Input image data is compressed by a first data compression unit 201 and stored in an external DRAM 103, and then decompressed by a first data decompression unit 202 and used for image coding. At the time of image coding, motion compensation data is generated using image data used at the time of past image coding, and a difference calculation unit 106 obtains a difference value. The difference value is decoded by the decoding unit 108, added to the motion compensation data, converted to decoded image data, compressed by the second data compression unit 204, and stored in the external DRAM 110. Further, the image data stored in the external DRAM 110 is expanded by the second data expansion unit 203 and used to generate future motion compensation data.
[Selection] Figure 1

Description

  The present invention relates to input image data and an image encoding apparatus that encodes and reduces the amount of data, and a method thereof.

  In recent years, with the advancement of digital signal processing technology and cost reduction, an image is sampled and processed as a digital signal. Therefore, the image is sampled at a predetermined period, and each sampled pixel and phoneme are converted into discrete sample values and handled. For example, the luminance signal of an HDTV broadcast signal is sampled at 74.25 MHz, and each pixel is often converted into an 8-bit sample value and handled. The input signal converted into the sample value in this manner is usually temporarily stored in the memory, and then read out from the memory as necessary, and the image quality enhancement process and the image encoding process are performed.

  In general, a moving image has a high correlation between frames or fields. Therefore, in image coding, MPEG coding or H.264 encoding is used. Like H.264 encoding, a difference value after correction (motion compensation) is obtained in accordance with motion between frames or fields, and this is encoded to greatly reduce the data amount. However, an HDTV broadcast signal or the like has a high processing operation frequency, a very high memory access speed, and a circuit cost and power consumption increase.

  Here, an example of reducing the memory access speed described in (Patent Document 1) will be described with reference to FIG.

  In FIG. 2, the image encoding apparatus includes an image input unit 101, a data compression unit 102, an external DRAM 103, a data expansion unit 104, a motion compensation unit 105, a difference calculation unit 106, an encoding unit 107, a decoding unit 108, and an addition operation. The unit 109 and the external DRAM 110 are configured with an output unit 111.

  The input image data input from the image input unit 101 is compressed by the data compression unit 102 and stored in the external DRAM 103. Thereafter, the compressed data is read out from the external DRAM 103 in an order suitable for encoding such as MPEG, and is decompressed by the data decompression unit 104 and used for image coding.

  Next, at the time of image encoding, motion compensation data is generated by the motion compensator 105 using image data that has been encoded and decoded in the past, and the compressed image data is compressed and expanded. Then, the difference calculation unit 106 obtains the difference value. The difference value is variable length encoded by the encoding unit 107 and output from the output unit 111. The output of the encoding unit 107 is decoded by the decoding unit 108, added to the motion compensation data, converted into decoded image data, and input to the external DRAM 110. Here, data obtained by decoding the image-encoded data is referred to as local decoded image data. The local decoded image data stored in the external DRAM 110 is used to generate future motion compensation data.

Since the conventional image encoding apparatus having such a configuration compresses image data at the input stage and stores it in the external DRAM, the memory access speed and the memory capacity for the external DRAM can be reduced.
JP-A-10-271516

  In the example of (Patent Document 1), as described in the document, data compression is performed on input image data. However, data obtained by decoding image-coded data (local decoded image) Data) is not compressed. This is because if the locally decoded image data is distorted by data compression, the distortion affects subsequent image coding, resulting in a large image quality degradation.

  However, since many memory accesses in actual image encoding processing are input / output of local decoded image data, there is a problem that if this cannot be compressed, the effect of reducing memory access and data transfer amount is reduced.

  In consideration of the above problems, the present invention enables image compression of local decoded image data, further reduces memory access and data transfer amount, and realizes a reduction in circuit cost and power consumption. The purpose is to provide.

  In order to solve the above problems and achieve the above object, according to the present invention, there is provided an image encoding device for encoding difference information between frames or fields, wherein the input image data is compressed using a first compression method. A first data decompression unit, a first data decompression unit that decompresses input image data compressed by the first data compression unit using a decompression method for the first compression method, and a storage unit that stores a compressed reference image; The compressed reference image is decompressed by using a decompression method for the second compression method, which is a compression method substantially equivalent to or less than that of the first compression method, the compression distortion of which is approximately equal to or less than that of the first compression method. A second data decompression unit, a motion compensation unit that generates motion compensation data obtained by performing motion compensation on the compressed reference image decompressed by the second data decompression unit, and the first data decompression unit. A difference calculation unit that generates a difference image between the input image data that has been input and the motion compensation data, a coding unit that performs variable length coding on the difference image, and a decoding that decodes the difference image that has been subjected to the variable length coding An adder that generates, as a reference image, an image obtained by adding the motion compensation data to the difference image decoded by the decoding unit, and a second compression method is used for the reference image And a second data compression unit that generates the compressed image as the compressed reference image.

  The present invention applies a compression method in which the first data compression method used for input image data and the second data compression method used for local decoded image data are almost the same in the configuration as described above. Further, the compression distortion by the second compression method is controlled so as to be slightly smaller than the compression distortion by the first compression method. With this configuration, distortion in the locally decoded image data can be minimized, and at the same time, even if new distortion occurs, the distortion is distorted to a value close to that of the input image data. Since it becomes invisible, the input / output data amount of the memory is reduced with almost no deterioration in image quality, the memory access transfer amount is greatly reduced, the circuit scale can be reduced, and the power consumption can be reduced.

  The present invention applies a compression method in which the first data compression method used for input image data and the second data compression method used for local decoded image data are almost the same method as described above. Further, the compression distortion by the second compression method is controlled so as to be slightly smaller than the compression distortion by the first compression method. Specifically, the quantization roughness used in the first compression method is equal to or coarser than the quantization used in the second compression method.

  In general, data that has been decompressed once by applying data compression can eliminate most of the redundancy corresponding to the data compression method. For this reason, even if data compression and expansion with the same configuration are performed again, almost no distortion occurs. This will be explained by using a specific example. When the input integer value is divided by 4 and rounded off by quantization of data compression, the distortion is -2, -1, 0, or +1. This means that compression distortion has occurred. In addition, when the quantized data is decoded, the divided values are multiplied by 4, so all values are multiples of 4. Next, once the data subjected to such quantization is subjected to data compression with an equivalent configuration again, a multiple value of 4 is input as the quantization input. It will not occur. That is, for data compression with the same configuration, distortion due to data compression hardly increases even if data compression is repeated twice or more.

  However, for local decoded image data, a new distortion is generated separately by the image encoding process. If this image encoding distortion is deformed by data compression, distortion is accumulated in the image encoding process and the image quality deteriorates. May be incurred.

  Therefore, in the present invention, for data compression on local decoded image data, the distortion caused by the data compression is equal to or smaller than the distortion generated by data compression on the input image. As a result, distortion in the locally decoded image data is minimized, and even if new distortion occurs, the distortion is distorted to a value close to that of the input image data, so that almost no deterioration in image quality is visible.

  As described above, the input / output data amount of the memory can be reduced with almost no deterioration in image quality, the memory access transfer amount can be greatly reduced, the circuit scale can be reduced, and the power consumption can be reduced.

  The best mode for carrying out the present invention will be described below with reference to the drawings.

(Embodiment 1)
FIG. 1 is a block diagram of the image coding apparatus according to the first embodiment. The same code | symbol is used for the part of the same structure as the conventional image coding apparatus. That is, the image input unit 101, external DRAM 103, motion compensation unit 105, difference calculation unit 106, encoding unit 107, decoding unit 108, addition calculation unit 109, external DRAM 110, and output unit 111 in FIG. The operation is almost the same as that in the encoding apparatus.

  In addition to these, the first embodiment of the present invention further includes a first data compression unit 201, a first data decompression unit 202, a second data decompression unit 203, and a second data compression unit 204.

  Input image data input from the image input unit 101 is compressed by the first data compression unit 201 and stored in the external DRAM 103. Thereafter, the compressed data is read from the external DRAM 103 in an order suitable for encoding such as MPEG, and is decompressed by the first data decompressing unit 202 and used for image coding. Input image data includes MPEG, H.264 H.264 image data.

  Next, at the time of image encoding, motion compensation data is generated by the motion compensator 105 using the image data that has been encoded and decoded in the past, and the input image data that has been compressed and expanded is used. The difference calculation unit 106 obtains a difference value. The difference value is variable length encoded by the encoding unit 107 and output from the output unit 111. The output of the encoding unit 107 is decoded by the decoding unit 108, added to the motion compensation data, converted into decoded image data (local decoded image data), and the second data compression unit 204 The data is compressed and input to the external DRAM 110. The local decoded image data stored in the external DRAM 110 is expanded by the second data expansion unit 203 and used for generating future motion compensation data.

  The compression method used in the second data compression unit 204 uses a method having a configuration substantially equivalent to the compression method used in the first data compression unit 201, and the quantization used in the compression of the second data compression unit 204 is the first data Quantization finer than the quantization used in the compression of the compression unit 201 is performed.

  With this configuration, in the example of FIG. 1, both the input image data and the local decoded image data can be compressed, so that the memory access transfer amount to the external DRAM can be greatly reduced. Also, since the first data compression, which is almost the same as the second data compression used for the local decoded image data, is performed on the input image, the data compression distortion to the local decoded image data is reduced, and even the data compression distortion occurs. However, since the input image data compressed and expanded by the first data compression is distorted to a value closer to that, as a result, almost no visual distortion is felt.

  Even in a simulation experiment result in which the present invention is actually applied, when the local decoded image data is compressed without using the first data compression, the number of continuous motion compensation frames (P pictures) increases by image coding. Although the distortion gradually increases, the effect that the large distortion is not detected even if the motion compensation frames are continuous by simultaneously applying the first data compression for the input image data and the second data compression for the local decoded image data. Is obtained.

  Note that the first data compression unit 201, the first data decompression unit 202, the data compression unit are controlled by controlling the encoding amount generated by the first compression method and the second compression method to a power of 2. 102, when the data decompression unit 104 accesses the external DRAM, the data transfer amount, data access, and data utilization become efficient.

  Although the present invention has been described above using the embodiments, the configuration of the image coding method used in the present invention can be implemented by configurations other than these. Further, the processing of the present invention can be realized by a processor using software, and can be implemented as a partial function of a large-scale integrated circuit.

  The image decoding apparatus of the present invention is useful as an apparatus mounted on a digital television, a digital camera, a digital video camera, a digital video recorder, and the like, and in particular, a broadcast standard and a recording system for which a transmission method is determined are determined. It is mounted on digital TVs, digital video cameras, digital video recorders, etc. that record AV data that receives and plays back high-bit-accuracy images that are transmitted and recorded while maintaining compatibility. Suitable for equipment.

Block diagram of image coding apparatus according to Embodiment 1 Block diagram of a conventional image encoding device

Explanation of symbols

101 Image input unit 103 External DRAM
105 Motion Compensation Unit 106 Difference Operation Unit 107 Encoding Unit 108 Decoding Unit 109 Addition Operation Unit 110 External DRAM
DESCRIPTION OF SYMBOLS 111 Output part 120 Image coding part 201 1st data compression part 202 1st data expansion part 203 2nd data expansion part 204 2nd data compression part 220 Image coding part

Claims (4)

An image encoding device that encodes difference information between frames or fields,
A first data compression unit that compresses input image data using a first compression method;
A first data decompression unit for decompressing input image data compressed by the first data compression means using a decompression method for the first compression method;
Storage means for storing the compressed reference image;
The compressed reference image is decompressed by using a decompression method for the second compression method, which is a compression method substantially equivalent to or less than that of the first compression method, the compression distortion of which is approximately equal to or less than that of the first compression method. A second data decompression unit;
A motion compensation unit that generates motion compensation data obtained by performing motion compensation on the compressed reference image decompressed by the second data decompression unit;
A difference calculation unit that generates a difference image between the input image data output from the first data decompression unit and the motion compensation data;
An encoding unit for variable-length encoding the difference image;
A decoding unit for decoding the variable length encoded difference image;
An addition operation unit that generates, as a reference image, an image obtained by adding the motion compensation data to the difference image decoded by the decoding unit;
An image encoding apparatus comprising: a second data compression unit configured to generate an image obtained by compressing the reference image using a second compression method as the compressed reference image. The first data compression unit and the first data expansion unit quantize quantization used in the first compression method to be equal to or coarser than quantization used in the second compression method. The image encoding apparatus according to claim 1. The first data compression unit controls the amount of data generated by the first compression method to be a power of approximately 2;
The image coding apparatus according to claim 1, wherein the second data compression unit controls the amount of data generated by the first compression method to be approximately a power of two. An image encoding method for encoding difference information between frames or fields,
First data means for compressing input image data using a first compression method;
First data means for decompressing the input image data compressed by the first data compression means using a decompression method for the first compression method;
Storage means for storing the compressed reference image;
The compressed reference image is decompressed by using a decompression method for the second compression method, which is a compression method substantially equivalent to or less than that of the first compression method, the compression distortion of which is approximately equal to or less than that of the first compression method. Second data decompression means;
Motion compensation means for generating motion compensation data obtained by performing motion compensation on the compressed reference image decompressed by the second data decompression means;
Difference calculation means for calculating a difference for generating a difference image between the input image data output from the first data decompression means and the motion compensation data;
Encoding means for variable-length encoding the difference image;
Decoding means for decoding the variable length encoded difference image;
An addition operation means for generating an image obtained by adding the motion compensation data as a reference image to the decoded difference image;
And a second data compression means for generating, as the compressed reference image, an image obtained by compressing the reference image using a second compression method.

Images