基于Cycle-spinning的多帧图像Contourlet去噪  

Cycle-spinning Based Contourlet Denoising for Multiple Image

作　　者：程燕[1] 

机构地区：[1]华东政法大学计算机科学与技术系,上海201620

出　　处：《计算机科学》2013年第11A期314-317,共4页Computer Science

基　　金：教育部人文社科项目(11YJCZH175)资助

摘　　要：本文提出一种改进的Cycle-spinning多帧Contourlet域图像去噪算法。根据视频序列连续帧间存在的运动信息,使用帧间位移矢量来代替平移技术对图像进行Contourlet去噪,基于帧间相关性不同的特点,改进Cycle-spinning变换为加权平均以消除伪吉布斯现象。实验结果显示该方法能有效去除各种类型的图像噪声,保留图像的细节和纹理信息,峰值信噪比(PSNR)有显著提高。A CS-based multiframe denoising algorithm in Contourlet domain is proposed. We exploit the state-space relationship among the sequential images to build a novel CS-based framework to reduce noise. By estimating state-space relationship among the inter-flames, the successive frames are denoised in Contourlet domain and used weighted average to get a restored image. Experiments demonstrate the effectiveness of the proposed method and show the superiority for various noise characteristics, especially in preserving the detail and texture information with higher PSNR values.

关 键 词：Cycle-spinning变换 CONTOURLET变换 多帧 图像去噪 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]