一种基于双树复小波变换的图像去噪算法

针对图像处理的需求,提出一种基于双树复小波变换的图像去噪算法。该算法对图像进行双树复小波变换,对变换后的系数利用最大似然估计和维纳滤波进行去噪,采用最大后验概率来估计双树复小波的方差,利用维纳滤波得到去噪后的系数,通过双树复小波反变换得到去噪后的图像。在分解层计算方差时,均采用在4×4的结构元素内基于最大值和次大值坍缩后的邻域来计算方差。实验结果表明,该算法的PSNR对比其他文献提高0.2 d B左右,运行时间减少5 s。

An Image Denoising Algorithm Based on Double-tree Complex Wavelet Transform

An algorithm for image denoising based on double-tree complex wavelet transform is proposed.The algorithm firstly applies double-tree complex wavelet transform to the images,and denoises the transformed coefficients using maximum likelihood estimation and Wiener filtering.The maximum posterior probability is used to estimate the variance of the double-tree complex wavelet,and Wiener filtering to get the denoised coefficient,finally the denoised image can be obtained through the double-tree complex wavelet inverse transform.The neighborhood in the4×4structural elements based on the maximal and the submaximal collapses is used to calculate the variance in the decomposition layer.The experimental results prove that the PSNR is improved by about0.2dB,and the operation time reduced by5s,as compared with other algorithms.