Abstract: Most image pre-processing approaches before encoding usually cause information loss. An adaptive and reversible image pre-processing method was proposed for a better performance of entropy coding. It was addressed by two steps, namely the bit-plane decomposition and the new reversible image transform scheme under different transform patterns in 8/4 pixel processing. Among transformed images, the one with the least entropy was selected as the pre-processing output with the histogram U-shape redistributed. The experimental results illustrate that the proposed method efficiently improves the performance of entropy coding by increasing the compression ratio than that of plain entropy coding.

Key words: data compression, entropy coding, image pre-processing, image transform, statistical characteristic, U-shape distribution

摘要： 大多数图像编码前所采用的预处理方法都会造成一定程度的信息损失。以提高图像的熵编码效率为目的，提出一种自适应可逆图像预处理方法——基于相邻像素相似性的可逆图像变换，它包括8像素和4像素处理下的多种模式。在对图像做位平面分解后进行基于相邻像素相似性的可逆图像变换，并选择多种变换方式下具有最小熵的变换图像作为预处理输出，其直方图呈现出U型分布。实验结果表明，该预处理方法改变了图像的统计特性，从而提高了图像的压缩比和熵编码效率。

关键词: 数据压缩, 熵编码, 图像预处理, 图像变换, 统计特性, U型分布