关键词: 小波变换, 图像去噪, 半软阈值法, Bayes估计

Abstract:

Based on the image denoising method of wavelet semisoft threshold, a novel method based on adaptive local correlation coefficient is proposed. By introducing the local correlation coefficient, the proposed method, which has a good compromise between the soft threshold and hard threshold, can enhance the subband correlation among wavelet coefficients in a variety of wavelet transforms. The adaptive threshold is selected based on Bayes risk estimation and the statistical significance in order to achieve the best estimation of wavelet coefficients in the different directions and subbands. Experimental results show that the new method improves the image denoising effect efficiently, and reduces the image visual distortion and edge oscillation caused by the image wavelet transform so as to retain the image feature of edges and detail at the same time. This method can be controlled by adjusting the partial correlation coefficients with the extent and effect of image denoising, thus meeting the different needs and having high practical value.

Key words: wavelet tansform;image denoising;semi-soft threshold;Bayes estimation