快速确定轮廓波消噪收敛因子

Fast Finding Shrinkage Factors for Contourlet Denoising

轮廓波变换能够有效地快速逼近图像信号,在图像消噪方面优于小波消噪.为了更好地保留图像消噪后图像的纹理信息和边界特征,一般采用硬阈值的消噪方法.本文分析了目前存在的3种硬阈值确定方法速度较慢的原因,通过较小样本图像确定每个子带的3σ或4σ收敛因子,在保持消噪效果(峰值信噪比和视觉效果)的同时,极大地提高了阈值确定速度,并大大减小了确定收敛因子所需的内存空间.

Contourlet transform can fast approximate to image signals efficiently, and its ability of image denoising is superior to that of wavelet. In order to retain borders character and texture information well in recovered image, hard threshold approach is often adopted. In this paper, we analyzed three different threshold methods and gave the reason that it is slow to find shrinkage factors. 3σ or 4σ shrinkage factors found through smaller sample images can improve the speed of finding threshold factors and reduce the memory required the shrinkage factors, meanwhile keep the denoising result not only in PSNR and visual quality.