全变分的分块小波阈值图像去噪  

Block wavelet threshold image denoising on total variation

作　　者：高媛媛[1] 刁永锋[1] 边赟[2] 

机构地区：[1]西华师范大学计算机学院,四川南充637002 [2]四川大学计算机学院,成都610064

出　　处：《计算机应用》2012年第5期1289-1292,共4页journal of Computer Applications

基　　金：教育部留学回国人员启动基金资助项目(20091341-11-3);四川省科技厅应用基础研究项目(05JY029-093)

摘　　要：针对全变分算法对平滑区域抑噪不充分及小波阈值算法易造成边缘模糊的缺陷,提出了全变分的分块小波阈值算法。首先利用全变分粗去噪算法对含噪图像去除幅度较大的噪声,再利用分块小波阈值算法抑制残余噪声。实验结果表明,在图像边缘得到保护的同时较好地抑制了平坦区域的残余噪声,与传统算法相比,峰值信噪比和视觉效果都得到明显提高。Traditional total variation algorithm of denoising could result in large residual noise in the smooth area and wavelet threshold algorithms could lead to some blurry edges. In view of these disadvantages, a new algorithm of block wavelet threshold image denosing on total variation was proposed. Firstly, this algorithm used total variation algorithm to denoise roughly, which eliminated noise in significant amplitude. Secondly, this algorithm used block wavelet threshold algorithm to eliminate the large residual noise in the smooth area. The experimental results show that the algorithm suppresses residual noise and protects edge effectively. Compared with traditional methods, both the peak signal-to-noise ratio and visual effects have been improved obviously.

关 键 词：全变分 分块 阈值去噪 模糊边缘 残余噪声 

分 类 号：TP394.41[自动化与计算机技术—计算机应用技术]