采用格形矢量量化的小波图象编码方法

Image coding method based on wavelet transform and lattice vector quantization

研究了以格形矢量量化方法进行小波图象的压缩编码。以两种经典的格形矢量量化算法乘积码塔形矢量量化（ＰＣＰＶＱ）和分块均匀格点矢量量化为例，分析了两种算法中非均匀矢量格点的分布与输入信号源的概率密度分布函数的关系，指出在保持矢量格点具有规则分布的前提下，格点分布难以与不规则的输入矢量概率分布实现良好的匹配。提出了一种均匀格点分布与熵编码相结合的矢量量化图象编码方法，该方法与以上两种算法对信号源输入矢量的概率分布具有更灵活的适应能力。给出了该算法和ＰＣＰＶＱ的实验结果的比较。

Wavelet Image coding methods based on lattice vector quantization (LVQ)) are studied. Taking product code pyramid veclor quantization (PCPVQ) and piecewise uniform LVQ as two examples, we analyse the relationship between non uniform lattice distribution and the source probability density function (PDF) and then point out a common mismatch problem of the two algorithms: with the regularity of lattice distribution, it is unrealistic for the lattice distribution to match the input vector distribution well. Meanwhile, a new image coding algorithm combining both uniform LVQ and entropy coding is put forward in this paper. The new algorithm can be adapted to the source PDF better than PCPVQ and piecewise uniform LVQ. Experimental results of both PCPVQ and the new algorithm are presented for comparison.