双变量模型与平移不变小波变换相结合的图像去噪算法

提出一种双变量模型与平移不变小波变换相结合的图像去噪算法.与传统的小波阈值去噪算法不同,该阈值不是直接作用于小波系数,而是利用图像平移不变小波变换系数尺度内和尺度间的双重相关性信息,构造出非高斯双变量分布模型,对该模型应用贝叶斯最大后验估值理论推导得到相应的非线性双变量阈值函数.同时,对第L级系数也作自适应处理.仿真结果和分析表明,该算法在视觉效果和峰值信噪比的改善上都取得了非常好的效果.

Algorithm of image de-noising combining translation invariant wavelet transform and bivariate mode

Based on the translation invariant wavelet transform and bivariate mode under the framework of Bayesian MAP estimate, this paper proposes a new algorithm of image de-noising. Unlike the traditional threshold shrink algorithm, our threshold relies not solely on the coefficients, but on the intra-scale and in- ter-scale correlations of TIWT coefficients,which we use to create non-Gaussian bivariate shrink distribute and the bivariate threshold function via estimate theory. Moreover the L resolution coefficients are man- aged by locally adaptive fashion. The experiment and analysis show an obvious improvement in both the subjective visual effect and the PSNR.