基于统计学原理的JPEG量化表的算法

JPEG图像压缩算法有较优秀的压缩品质，但在量化阶段引用的是心理视觉阀值构成的量化表，造成了大量具有统计意义的信息的损失，给还原后的图像识别带来了困难。为此，本文提出了一种新的量化表设计方案，并运用数理统计理论推导出判别最优划分的标准，进而导出量化表的算法和实现步骤。此算法还可广泛地应用于非均匀量化的A／D、D／A转换等方面。

Statistics-based Arithmetic of JPEG Quantization Table

Arithmetic of JPEG image compressing is of high compressing quality. At the quantizationstage it employs quantizationtable formed by psychological vision thresholdand damages a lot of statistical information , leading to difficulty of recognising the reverted images .Therefore, a new design of quantizationtable was put forward in this paper. We deduce the measuring criteriaof recognising the optimal division and the arithmetic and process based on statistical theory. This arithmetic can be applied widely with non-averagely quantizationA/D, D/A conversion, etc.