| 摘要 |
The present invention relates to a method and apparatus for compacting gray-scale image data which maintains extremely good picture quality and can result in typical images in a compression ratio of 5 to 1 and even better if a final step of arithmetic coding is used. The method is a special application of adaptive differential pulse code modulation (DPCM) which utilizes a prediction of smoothness surrounding a current pel to determine variable length codes to convey the quantized error. Thus, a continuously adaptive variable length code is produced which may be accurately decoded without using any further marker (code indicator) bits. For each pel and "error bit" is generated to indicate to the decoder if the predicted quantization range is adequate. These eroror bits can be further significantly compressed. This method permits the use of a "zero bit" per pel mode for the data transmission (1 "error bit" per pel actually transmitted) and provides for the automatic detection of the particular variable length compression code utilized for transmitting the quantized error (EY) for a particular pel. Utilizing the concepts of the present invention the inventors have obtained compression rates of approximately 1.5 bits per pel that produce a gray scale image at the decoding end of such a data transmission or storage system which has excellent quality (equal to 5 bit DPCM). By adding the concept of arithmetic coding to the previously set forth method, due to the nature of the statistical distribution of said special purpose error bit patterns, a compressed data set of less than 1 bit per pel has been achieved. This was produced from an original 8 bit per pel gray scale image. |
