论文部分内容阅读
本文研究一种静态图像小波（Ｗａｖｅｌｅｔ）域系数压缩方法，可以在任意给定比特数时获得一定视觉意义下的最佳图像质量。该方法根据图像小波分解和人类视觉（ＨＶＳ）的特点及其关系，对系数进行不同间隔的量化，并规定系数的视觉重要性顺序，结合零树数据结构，对系数进行该顺序的比特层零树扫描和预测，输出符号数据流，最后用自适应算术熵编码实现高效率编码。该方法易于用ＶＬＳＩ实现，输出比特率任意可调。计算机模拟结果显示，对于图像（Ｌｅｎａ２５６）在０．２～０．３ｂｉｔ／ｐｉｘｅｌ时仍可获得较满意的重构图像质量
In this paper, we study a method for compressing coefficients in a wavelet domain of static images, which can obtain the optimal image quality in a certain visual sense at any given bit number. According to the characteristics of wavelet transform and human visual system (HVS) and their relations, this method quantizes coefficients at different intervals and prescribes the order of visual importance of coefficients. Combining with the zero-tree data structure, Tree scanning and prediction, the output symbol data stream, and finally with adaptive arithmetic entropy coding to achieve high efficiency coding. The method is easy to implement with VLSI and the output bit rate can be adjusted freely. Computer simulation results show that for the image (Lena256) at 0.2 ~ 0.3bit / pixel can still get more satisfactory reconstructed image quality