一种基于小波分解的图象分层分类矢量量化方法

A Classified Vector Quantization Scheme of Still Image Based on Wavelet Transform

作者提出了一种基于小波分解及采用自组织特征映射神经网络进行分层分类矢量量化的静态图像压缩编码方法。首先对图象进行小波分解。利用不同分辩率级间小波子图的相似性 ,将最低分辩率层子图的矢量编码信息作为整幅图象的解码数据 ,并将其矢量分类和解码索引地址信息用于高分辩率层子图的码书训练。实验表明 ,和其他文献提出的方法相比 ,作者提出的方法在获得较好重构图象质量的前提下 。

In this paper, a new image coding scheme based on wavelet transform and Hierrchical-classified vector quantization (HVCQ) using Self Organizing Feature Map(SOFM) neural network is presented. The image was decomposed by discrete wavelet transform. The class information was self generated according to similarity between subimages. The evctor quantization coding information of lowest resolution subimage was only used for decoding the whole image, the class information and decoding index address was used for the vector codebook training of the high resolution subimage. Sompared with the published schemes, The experiment results show that this new scheme perform better in the aspect of compression ration and decoding speed under good restored image quality.