• 专利附录
  • 专利说明
  • 权利要求
  • 类似技术
  • 同族专利
  • 引用文献
  • 法律状态
  • 优先权
    • 专利摘要:
    The present invention relates to an image compression method and apparatus using variable quantization encoding that can significantly reduce the loss due to quantization error occurring in the conventional quantization process using the variable quantization technique. The present invention includes a motion detector (1) for extracting a motion vector of the front and rear screen; A frame memory section 3 for accumulating a previous screen; A correlation comparison unit (2) (4) for comparing the motion vectors of the front and rear screens and the front and rear screens to define the degree of correlation; A reset signal timing generator 5 for generating reset signal timing for varying the quantization matrix value according to the correlation comparison result; An image compression apparatus using a variable quantum encoding composed of a variable quantization matrix unit 6 for varying quantization matrix values, wherein the correlation comparison unit compares the motion vectors of the front screen A and the rear screen B with each screen. The quantization value of the screen stored in the memory is maintained as the quantization matrix value until the comparison value of the correlation becomes A> B, and the comparison value of the front and rear screens (A) (B) is A <B. When the reset timing signal is generated, the rear screen B is applied to the memory and the quantization matrix value that has been maintained is reset, and the quantization matrix value of the rear screen B applied to the memory is applied to the quantization matrix and the rear screen ( It is an image compression method using a variable quantum coding that maintains a quantization matrix value until the comparison signal of B) becomes B <C.
    公开号:KR19990036476A
    申请号:KR1019980015199
    申请日:1998-04-28
    公开日:1999-05-25
    发明作者:장진혁
    申请人:백정한；
    IPC主号:
    专利说明:

    Image Compression Method Using Variable Quantum Coding and Its Apparatus
    The present invention relates to an image compression method and apparatus using variable quantum coding used for image compression such as a moving picture, and more particularly, to a variable quantization technique to dramatically reduce the loss due to quantization error occurring in the existing quantization process. The present invention relates to an image compression method and apparatus using variable quantum encoding that can provide a high-definition, high-definition, high-rate compressed image by reducing
    The conventional image compression method of applying discrete cosine transform (DCT) to image compression such as a moving image in image compression such as a moving image is performed after A / D conversion (sampling and quantization) in an analog source, as shown in FIG. Data was compressed by encoding information using a DCT (Discrete Cosine Transform).
    The encoding of information is roughly classified into a DCT transform and an entropy encoding. The DCT transform as shown in FIG. 2 is a kind of orthogonal transform that transforms an image from a spatial coordinate to a frequency.
    In general, as shown in FIG. 3, a natural image has a small change between adjacent pixels and a high correlation between pixels.
    This means that the spatial frequency expressed as the number of conversions per unit distance is low.
    Therefore, when a frequency component is decomposed by an orthogonal transformation of an image having such a correlated image, a low frequency component has a large value, while a high frequency component is close to 'O'.
    As a result, the frequency component distribution after the conversion has a bias with the DC component as the vertex, and when combined with the entropy coding, the code amount can be reduced.
    The process of removing the correlation of an image using the characteristics of such an image signal, such as a still image, moving picture, a secondary image, a luminance signal, a chrominance signal, or converting the image using a visual characteristic to easily reduce redundancy. Modeling.
    In the modeling process, the number of transformed and output data is the same as the input digital image, and the informative entropy becomes small, thereby facilitating the reduction of data through entropy encoding.
    Entropy encoding is to reduce the amount of code to be transmitted or accumulated by allocating an appropriate code using the statistical properties of the image signal converted in the modeling step of encoding the image information source, that is, to compress the data.
    In general, as mentioned above, a natural image has a high correlation between adjacent pixels.
    However, a prediction encoding technique as illustrated in FIG. 4 is commonly used to encode entropy using such a correlation of images.
    The predictive encoding technique is an encoding scheme that converts an image signal into a signal with low entropy by using the correlation of the image.
    In other words, the pixel value to be encoded is predicted from the encoded and decoded pixel values to encode an error with the original pixel value.
    That is, the encoding is performed by using the property that the entropy of the prediction error signal decreases as the predictions match.
    Referring back to FIG. 1, the conventional image compression tool first undergoes an A / D conversion process.
    This is necessary to perform the vast calculations necessary for compression with minimal degradation of the signal since most of the existing image sources are analog signal sources.
    After the analog signal has undergone digital conversion, the signal passes through an encoder as shown in FIG.
    In the case of a moving picture, the encoder is delivered to a detection stage that detects motion as a vector value, and detects the motion by using the correlation of the signal, and applies the detected signal to a frame memory, that is, a predictor.
    The signal applied to the predictor becomes the basis of the prediction mode and is used as a compression tool using the characteristics of assimilation.
    The next step is a part which is inevitably passed regardless of a still image or a moving picture, and goes through a discrete cosine transform (DCT) block, which is an orthogonal transform that transforms spatial coordinates into frequency coordinates.
    This can be said to be a modeling block for reducing the entropy during entropy coding by having a bias of the DC component as a peak.
    The video signal passed through this modeling block passes through a quantization block, which is the next step, and the data is greatly reduced during the quantization process.
    In FIG. 6, a typical quantization process uses a quantization matrix called a quantization matrix, and the DC component of the intra block variably encodes the difference value with all the blocks, and the AC component of the intra block and the quantization coefficient of the non-intra block are Variable-length coding is performed by setting the number of consecutive runes of the value 'O' and the value (level) of the coefficients other than 'O' as a pair.
    As shown in Fig. 7, these encodings are performed by zigzag scanning from low frequency components to high frequency components.
    Since the motion vector is also efficiently encoded, the difference value of the vector is variable-length coded, and information such as the variable-length coded discrete cosine transform (DCT) coefficient is generated at an irregular rate but is accumulated at a constant rate and is thus stored in the buffer.
    Here, the coding is controlled by changing the quantization scale factor q corresponding to the code amount stored in the buffer in order to make the average of the coding rates constant.
    For example, if the code amount is too large, the value of q is increased so that a large number of coefficients 'O' is generated, and if the code amount is small, q is reduced.
    As such, the conventional compression tool is obtained in the quantization process by greatly reducing the data, but it has been pointed out that the problem is that the ratio of reduction is determined by the quantization error occurring in the quantization process.
    Since the quantization process needs to be quantized by a quantization matrix, which is a predetermined quantization matrix, the discrete cosine transformed coefficients using the predetermined quantization matrix are matched with a target value to reduce the scale of the quantization matrix. This is because the division was made to increase the occurrence of O '.
    In other words, when a large compression ratio is applied, the fine signal of the original image is canceled out to restore a coarse image having a reduced resolution.
    As described above, it is inevitable that image compression is greatly inevitable in conventional compression tools, and it is unreasonable to perform a significant compression while maintaining the basic resolution.
    In addition, since the existing compression tools perform compression within a range that does not cause deterioration of the basic resolution, degradation of resolution inevitably occurs when the compression ratio exceeds the specified compression ratio.
    The present invention was devised to solve the disadvantages of the conventional compression tool as described above.
    The present invention has a quantization matrix that can easily change the quantization matrix consisting of a fixed quantization matrix in the quantization process by using image correlation to have a much higher compression ratio than the existing compression tools. The reset timing of the quantization matrix value is extracted by using the correlation vector of the motion vector through the digital comparator and the correlation between the front and rear screens accumulated in the frame memory. The quantization matrix is constructed, and the quantization matrix value is maintained until the reset timing signal is generated according to the degree of correlation, and when the reset timing signal is generated, the quantization matrix value is reset to reset the quantization matrix value. Using as a quantization matrix as is An image compression device using variable quantum coding was constructed to reduce the loss due to quantization error during the compression process so that the image can be compressed at a high compression rate.
    1 is a flowchart of general information source encoding (compression and restoration).
    2 is a diagram illustrating encoding by DCT (discrete cosine transform)
    3 is an explanatory diagram illustrating the correlation of image signals
    4 is a flowchart of predictive encoding of entropy encoding using correlation of image signals.
    5 is a diagram illustrating an encoder mounted after an analog signal undergoes digital conversion.
    6 is a quantization matrix diagram in a quantization process
    7 is an illustration of scan (in zigzag) by DCT
    8 is a block diagram of an image compression device using variable quantum encoding according to the present invention.
    <Description of Symbols for Main Parts of Drawings>
    1: motion detection unit 2, 4: correlation comparison unit
    3: frame memory section 5: reset signal timing generator
    6 variable quantization matrix
    In Fig. 8, the image compression device using the variable quantum encoding according to the present invention comprises: a motion detection unit 1 for extracting motion vectors of the front and rear screens; A frame memory section 3 for accumulating a previous screen; A correlation comparison unit (2) (4) for comparing the front and rear screens with their motion vectors to define a degree of correlation; A reset signal timing generator 5 for generating reset signal timing for varying the quantization matrix value as a result of the correlation comparison; It consists of the variable quantum matrix part 6 which changes a quantization matrix value.
    Here, suppose the front screen is A, the rear screen is B, and the comparison screen of A and B is C.
    As a result of comparing the motion vector with the front and rear screens A and B, the quantization value of the screen stored in the memory is maintained as the quantization matrix value until the correlation becomes A> B.
    When the comparison value of the front screen A and the rear screen B is A <B, the reset timing signal drops and the rear screen B is applied to the memory.
    At this time, the quantization matrix value that has been maintained is reset, and the quantization matrix value of the rear screen B is applied to the memory and then applied to the quantization matrix, and is maintained until the comparison signal of the rear screen B becomes B <C. .
    As in the conventional image compression tool, the quantization matrix itself is already defined, and an error occurs (quantization error) due to variable tuning of the quantization scale according to the compression rate in the quantization process, resulting in deterioration of the signal. By going through the quantization process using the signal degradation is significantly reduced.
    Here, the degradation of the signal due to the reduction of the quantization error will be described in detail.
    The memory and the quantization matrix values are reset or maintained according to the comparison signal generated by the comparator.
    The signal applied to the memory has its own quantization value.
    Therefore, if the quantization matrix is constructed by substituting this unique quantization value into the quantization matrix value of the quantization matrix, quantization is performed according to the correlation of signals in the quantization process, and thus a high compression ratio can be obtained.
    In other words, a natural image has a very high correlation between pixels between pixels, lines between lines, fields and fields, frames and frames.
    This means that the change in screen is not that big.
    Therefore, if the quantization value of the front screen A is divided by the quantization value of the rear screen B by using the property of highly correlated images, the correlation part is canceled and the coefficient of 'O' is generated. Bits are added, and uncorrelated portions remain in error and long bits are added by variable length coding.
    This means that there are more coefficients that become 'O' due to signal correlation than coefficients that do not become 'O'.
    This is because it is proportional to the output of the previous comparator.
    When the output of the correlation comparison unit 2 (4) compares the current picture with the subsequent picture and the subsequent picture shows a larger value (indicating no correlation) than the current picture, a reset signal of the memory and the quantization matrix value is generated. At the point where the quantization matrix is maintained all the time, the coefficient of 'O' must increase, and if the frequency of occurrence of the 'O' coefficient increases during this quantization process, short bits can be allocated in the variable length coding process, so fixed quantum It is much more efficient to code than using quantization matrices with matrices.
    In sum, the motion detector extracts the motion vectors of the front and rear screens. A frame memory unit for accumulating a front screen; A correlation comparison unit for comparing the front and rear screens with their motion vectors to define a correlation degree; A reset signal timing generator for generating reset signal timing for varying the quantization matrix value as a result of the correlation comparison; The image compression apparatus of the present invention organically connecting a variable quantum matrix unit for varying quantization matrix values has the following effects.
    Unlike using a fixed and prescribed quantization matrix like a conventional image compression tool, a variable quantization matrix can be configured as the quantization matrix is automatically changed according to the correlation of the image, so that a finer and clearer high-definition image can be obtained. have.
    Field of the Invention The present invention is a field requiring high-definition signal recovery, such as PACS, digital TV, G4 fax, high-definition broadcasting system, and other fields requiring high-quality images. It is effective in the field where the compression ratio is required, such as digital cable broadcasting and airwave broadcasting, G4 fax and wired and wireless image data transmission.
    权利要求:
    Claims (5)
    [1" claim-type="Currently amended] A motion detector 1 for extracting motion vectors of the front and rear screens; A frame memory section 3 for accumulating a previous screen; A correlation comparison unit (2) (4) for comparing the motion vectors of the front and rear screens and the front and rear screens to define the degree of correlation; A reset signal timing generator 5 for generating reset signal timing for varying the quantization matrix value according to the correlation comparison result; And a variable quantization matrix section (6) for varying quantization matrix values.
    [2" claim-type="Currently amended] A motion detector for extracting motion vectors of the front and rear screens; A frame memory unit for accumulating front screens; A correlation comparison unit for comparing the motion vectors of the front and rear screens and the front and rear screens to define a degree of correlation; A reset signal timing generator for generating reset signal timing for varying the quantization matrix value according to the correlation comparison result; In a variable quantum encoding apparatus including a variable quantum matrix unit for varying a quantization matrix value, the correlation comparison unit compares each screen with a motion vector of the front screen A and the rear screen B and the comparison value of the correlation is A>. The quantization value of the screen stored in the memory is maintained as the quantization matrix value until the value B becomes, and the reset timing signal is generated when the comparison value of the front and rear screens A and B becomes A <B. The screen B is applied to the memory, and the quantization matrix value, which has been kept, is reset and the quantization matrix value of the rear screen B applied to the memory is applied to the quantization matrix so that the comparison signal of the rear screen B is B <C. An image compression method using a variable quantum encoding, wherein the image compression is performed while maintaining the quantization matrix value until the quantization matrix value is maintained.
    [3" claim-type="Currently amended] The method of claim 2, wherein the correlation comparison unit resets or maintains the memory and the quantization matrix value according to the comparison signal, and the signal applied to the memory has a unique quantization value, and the quantization value is converted into a quantization matrix value of the quantization matrix. And a quantization matrix formed by substituting the quantization matrix so that the quantization is performed according to the signal correlation in the quantization process so that the compression of the image is performed.
    [4" claim-type="Currently amended] 3. The reset signal of the memory and the quantization matrix value is generated if the output of the correlation comparison section compares the image currently held with the subsequent image and the subsequent image is larger than the image currently held (no correlation). An image compression method using variable quantum coding, characterized in that the compression of an image is performed by determining that the occurrence rate of the 'O' coefficient is high when the quantization matrix is maintained.
    [5" claim-type="Currently amended] The quantized value of the previous screen is divided by the quantized value of the subsequent screen by using properties of a highly correlated image to cancel the correlated portion, and an 'O' coefficient is generated to generate variable length coding (adding short bits). ), The uncorrelated part remains as an error and is variable length coded (long bit is added). And compressing the image so as to perform image compression.
    类似技术:
    公开号 | 公开日 | 专利标题
    JP2016167831A|2016-09-15|Method for use in coding system of block conversion base
    EP2202989B1|2014-04-16|Image encoding apparatus, image decoding apparatus, image encoding method, image decoding method, image encoding program, and image decoding program
    KR100371129B1|2003-02-07|Image predictive decoding apparatus and method thereof
    US7974346B2|2011-07-05|System and method for generating video data for implicit coefficient prediction decoding
    US6895050B2|2005-05-17|Apparatus and method for allocating bits temporaly between frames in a coding system
    US7782956B2|2010-08-24|Signal compressing system
    KR100468844B1|2005-01-29|Optimal scanning method for transform coefficients in image and video coding/decoding
    CA2295689C|2003-12-09|Apparatus and method for object based rate control in a coding system
    US5486862A|1996-01-23|Motion picture encoding system
    US7010035B2|2006-03-07|Quantization matrix for still and moving picture coding
    KR0155784B1|1998-12-15|Adaptable variable coder/decoder method of image data
    JP3743960B2|2006-02-08|Image decoding method
    CN100581252C|2010-01-13|Method for improving entropy coding
    US7460723B2|2008-12-02|Quality based image compression
    AU782176B2|2005-07-07|Variance based adaptive block size DCT image compression
    US6658157B1|2003-12-02|Method and apparatus for converting image information
    KR100527690B1|2005-11-15|Signal encoding method and apparatus and decoding method and apparatus
    JP2675529B2|1997-11-12|Video coding method and apparatus thereof
    JP3678482B2|2005-08-03|Video data post-processing method
    DE60309375T2|2007-02-15|Parameterization for course compensation
    US6757438B2|2004-06-29|Method and apparatus for video compression using microwavelets
    JP3442028B2|2003-09-02|Data decoding method and apparatus
    US5748121A|1998-05-05|Generation of huffman tables for signal encoding
    KR960010196B1|1996-07-26|Dct coefficient quantizer utilizing human vision characteristics
    JP2509469B2|1996-06-19|Method and apparatus for adaptive encoding and decoding of continuous image by conversion
    同族专利:
    公开号 | 公开日
    引用文献:
    公开号 | 申请日 | 公开日 | 申请人 | 专利标题
    法律状态:
    1998-04-28|Application filed by 백정한
    1998-04-28|Priority to KR1019980015199A
    1999-05-25|Publication of KR19990036476A
    优先权:
    申请号 | 申请日 | 专利标题
    KR1019980015199A|KR19990036476A|1998-04-28|1998-04-28|Image Compression Method Using Variable Quantum Coding and Its Apparatus|
    [返回顶部]