The discrete cosine transform (DCT) has been shown as an optimum encoder for sharp edges in an image (Andrew and Ogunbona, 1997). A conventional lossless coder employing differential pulse code modulation (DPCM) suffers from significant deficiencies in regions of discontinuity, because the simple model cannot capture the edge information. This problem can be partially solved by partitioning the image into blocks that are supposedly statistically stationary. A hybrid lossless adaptive DPCM (ADPCM)/DCT coder is presented, in which the edge blocks are encoded with DCT, and ADPCM is used for the non-edge blocks. The proposed scheme divides each input image into small blocks and classifies them, using shape vector quantisation (VQ), as either edge or smooth. The edge blocks are further vector quantised, and the side information of the coefficient matrix is saved through the shape-VQ index. Evaluation of the compression performance of the proposed method reveals its superiority over other lossless coders.
展开▼
机译：离散余弦变换（DCT）已被证明是图像中锐利边缘的最佳编码器（Andrew and Ogunbona，1997）。由于差分模型的简单模型无法捕获边缘信息，因此采用差分脉冲编码调制（DPCM）的传统无损编码器在不连续区域中存在明显的缺陷。通过将图像划分为据认为统计上固定的块，可以部分解决此问题。提出了一种混合无损自适应DPCM（ADPCM）/ DCT编码器，其中边缘块用DCT编码，而ADPCM用于非边缘块。提出的方案将每个输入图像分为小块，并使用形状矢量量化（VQ）将它们分类为边缘或平滑。进一步对边缘块进行矢量量化，并通过shape-VQ索引保存系数矩阵的边信息。对所提出方法的压缩性能的评估显示出其优于其他无损编码器的优越性。
展开▼
