关键词: 图像去噪, 曲波阈值法, 卡方分布, 偏微分方程

Abstract:

To circumvent the visual distortion due to the discontinuity of the hard threshold function and the constant reconstruction deviation caused by the soft thresholding function, this paper presents a novel thresholding function based on the Chi-Square cumulative distribution function according to the distribution characteristics of curvelet coefficients of the noise and the ideal properties thehigh effective curvelet threshold functions should have. Further, in order to eliminate the surrounding effect inherent in curvelet threshold denoising methods and achieve a better balance between detail conservation and noise reduction, useful information involved in a denoised image produced by the partial differential equation denoising method is fused with that by the novel curvelet threshold function into the proposed denoising method. Theoretical analysis and simulation results show that the proposed denoiding method outper forms the soft and hard threshold denoising methods in terms of the denoising effect and visual quality.

Key words: image denoising, curvelet thresholding algorithm, Chi-Square distribution, partial differential equation