一种自适应多尺度积阈值的图像去噪算法

该文提出了平稳小波变换(Stationary Wavelet Transform, SWT )域自适应多尺度积阈值的图像去噪算法(SWT domain Multiscale Products, SWTMP)。与传统的阈值去噪算法不同，该阈值不是直接作用于小波系数，而是作用于小波系数的空间多尺度积。分析了SWT域含噪图像多尺度积的特点，提出了SWT域自适应多尺度积阈值的计算方法。多尺度积强化了图像的重要结构信息，弱化了噪声，在有效去噪的同时更多地保留了图像的边缘和细节。实验结果表明，所提算法对自然图像去噪后的视觉效果和性能指标均好于二进小波域多尺度积阈值(Adaptive Multiscale Products Thresholding, AMPT)去噪方法。

An Image Denoising Algorithm Using Adaptive Multiscale Products Thresholding

An adaptive thresholding algorithm for natural image denoising is proposed in this paper, which is based on Stationary Wavelet Transform (SWT) and multiscale products. Different from traditional thresholding denoising algorithm, this threshold is imposed on multiscale products instead of imposed on wavelet coefficients directly. The characteristics of multiscale products of noisy image in SWT domain are analyzed, and an adaptive threshold estimator is proposed. The multiscale products intensify the important structure information of images and weaken the noise, and reach the result of both effective denoising and preserving the edges and details of image simultaneously. Experimental results show that the visual effect and the performance index of proposed algorithm outperform the adaptive multiscale products threshold denoising in dyadic wavelet domain.