对图像压缩编码算法进行了改进。首先,将小波分解后的3个高频系数进行预处理:将高频部分进行球坐标变换,降低了同一尺度内系数的相关性;基于小波域和球坐标域的两个前提,定义了多尺度模积的概念,用来控制收缩函数对小波高频部分进行收缩处理。这样,可以去除那些不影响视觉效果的小波系数以及噪声信息,达到较高的压缩比。然后,对小波变换的低频部分进行单独编码(DPCM),对球坐标下的高频部分采用改进的多级树集合分裂(SPIHT)编码。针对SPIHT编码中重复扫描的问题,引入了最大值矩阵MMP(matrix of maximum pixel),这种策略能够有效降低比较次数。仿真实验表明,本文提出的算法具有较好的编码效率。
Abstract:
In this paper,the image compression encoding algorithm is improved.The first,the high-frequency coefficients are transformed to the spherical domain to reduce the correlation in the same scale.Then,based on two premises,the multi-scale model product is defined,which is used to control the shrinkage function.In this way,the small wavelet coefficients and some noise can be removed.This strategy ensures a higher compression ratio.The low-frequency part is coded by DPCM,the high-frequency parts in spherical domain are coded by the improved set partitioning in hierachical trees(SPIHT) algorithm.In the new algorithm,the matrix of maximum pixel(MMP) is introduced to overcome the disadvatage of the traditional SPIHT algorithm.Experimetal results indicate that the new algorithm is effective and improves the images signal to noise ratio.