| 摘要 |
A DCT based lossy compression method for compressing a digitized image composed of a matrix of image samples to provide a compressed image which satisfies a predefined bit budget. The digitized image is first sub-divided into blocks (e.g. of size 8x8 pixels). A discrete cosine transform (DCT) comprising a set of DCT coefficients is then derived for each block. A quantization table is selected from a set of quantization tables and, using the selected table, the coefficients of each DCT are quantized. A zero-value index, corresponding to the average number of zero value quantized DCT coefficients per DCT, is determined. A predicted zero-value index is calculated using said predefined bit budget and a quantization table selected from said set of tables using the determined index and the predicted index. Using that selected table, the unquantized coefficients of the DCTs are re-quantized and the requantized coefficients compressed using run-length encoding and Huffman encoding. |
