摘要
为了提高对纹理丰富图像的压缩效果,首先用小波包变换对纹理丰富的图像进行完全分解,再用一种与后续编码器相关联的代价函数进行最佳基搜索,以提高对纹理图像的表达能力.为了提高编码算法的率-失真性能,对搜索出来的最佳基所对应的小波包系数采用变块尺寸方法进行分类,再对分类结果采用网格编码量化器进行量化,最后对量化结果进行熵编码.仿真结果表明,所构造的编码器与SPIHT算法相比,峰值信噪比提高了0 5～2dB,可用于对指纹、遥感等图像的压缩.
In order to improve the compression efficiency of texturerich images, wavelet packet transform is firstly applied to texturerich images for a full decomposition, then a cost fuction related with the sequential coding scheme is utilized for best basis selection to improve the representation efficiency of texturerich images. To improve the ratedistortion performance, a variable block size classification technique is used to classify the wavelet packet coefficicnts corresponding to the selected best basis, the trellis coded quantization then is used to quantize the classification results, and the quantization outputs are supplied to the entropy coder. Simulation shows that the proposed coder improves the peak singnaltonoise ratio by 05 to 2 dB for texturerich images when compared with the SPIHT coder.
出处
《西安交通大学学报》
EI
CAS
CSCD
北大核心
2002年第12期1249-1252,1294,共5页
Journal of Xi'an Jiaotong University
基金
教育部博士点基金资助项目(2000068928)
教育部骨干教师基金(2000年度)资助项目
西安交通大学重点培植项目
西安交通大学自然科学基金资助项目.
