Avoiding accumulation of deterioration caused by iterated dct image coding
| Autor: | Dai Ando, Kazuhisa Yanaka, Suguru Higashino |
|---|---|
| Rok vydání: | 1993 |
| Předmět: |
Computer Networks and Communications
ComputingMethodologies_IMAGEPROCESSINGANDCOMPUTERVISION Variable-length code Data_CODINGANDINFORMATIONTHEORY computer.file_format Coding tree unit JPEG Sub-band coding Shannon–Fano coding Computer Science::Computer Vision and Pattern Recognition Electrical and Electronic Engineering Arithmetic Quantization (image processing) Algorithm computer Decoding methods Mathematics Image compression |
| Zdroj: | Electronics and Communications in Japan (Part I: Communications). 76:1-11 |
| ISSN: | 1520-6424 8756-6621 |
| DOI: | 10.1002/ecja.4410760601 |
| Popis: | Discrete coding transform (DCT) coding is a highly efficient coding method for still images with continuous intensity. It has a feature that a drastic image compression can be realized. Among the coding methods considered by JPEG for international standardization, it is adopted as the most basic baseline system which is a cooperative organization of CCITT and ISO. However, the method is an irreversible coding. When the coding/decoding is applied iteratively to the image once coded and decoded, the image deterioration progresses due to the accumulation of the error. The major causes of the error accumulation are the quantization of the transform coefficient after DCT and the rounding off of the value when the intensity is converted to the integer value in the decoding. This paper shows theoretically the following property. When the quantization step Q satisfies a certain condition and if the image after coding/decoding stays within the range that can be represented by the same number of bits as the original image, no further image deterioration takes place in the following sense. When the decoded image is encoded again using the same quantization table, the result is the same as the previous value. When the image after coding/decoding is outside the range that can be represented by the number of bits of the original image, the value must be restricted. Thus, it is not necessarily true that the DCT transform coefficient of the image after encoding the image is the same as in the previous image. However, it is seen by the result of simulation that even in such a case, the process converges by a finite number of iterative codings, with only a slight deterioration of the signal-to-noise (SN) ratio. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|