基于Shearlet变换域图像噪声混合滤波方法

在深入分析Shearlet变换理论的基础上,给出具体的Shearlet变换步骤,并提出一种针对图像脉冲噪声的混合滤波方法.该方法首先对噪声图像进行Shearlet多尺度变换,获得高频和低频系数,保留低频系数不变;然后针对Shearlet高频系数能够较好地刻画噪声局部化的性质,设计出一种多尺度、多方向特性的串/并联自适应中值滤波器,对高频系数中噪声进行自适应检测和滤除;最后实现低频系数和滤波后高频系数重构.通过与小波阈值法去噪和开关中值滤波法仿真实验比较可知,该滤波方法对于高强度的脉冲噪声具有较好的处理效果.

A new hybrid filtering algorithm for image noise based on Shearlet transform domain

The theory on Shearlet transform is in-depth analyzed,and specific steps of Shearlet transform is put forward,a new hybrid filtering algorithm is given so as to deal with the salt & pepper nois in image.Firstly,the noise image is conducted Shearlet transform,obtaining high-frequency and low-frequency coefficient.Duo to the low-frequency is not bothered by image noise,so it is being keep unchanged.Secondly,the high Shearlet transform high-frequency coefficient can characterization the localization properties of the noise,so a new adaptive median filter is designed.the filter has the characteristic of multi-scale,multi-direction,series/parallel to detect and filter the image noise effectively. Finally,the low-frequency coefficient and high-frequency coefficient is reconstructed.The experiment results show that,the filtering method in this paper is better for high intensity salt& pepper noise than the Wavelet threshold and adaptive switch median filtering method.