基于改进阈值函数的自适应图像去噪方法

Adaptive Image Denoising Based on the Improved Threshold Function

经典的BayesShrink通过软阈值函数对小波系数进行修正,使得原小波系数和修正小波系数存在恒定偏差,降低了去噪图像质量。文章利用噪声系数幅值随小波系数幅值的增大而相对减小的特点,构造了具有伸缩性的自适应阈值函数,该方法克服了软阈值函数产生的恒定偏差,依赖小波系数幅值对原始系数和修正系数间的恒定偏差自适应调节。伸缩因子m根据不同图像特点采用二分法自适应寻得最优值。实验结果表明改进后的阈值函数相比经典软阈值函数去噪效果较优。

Bayes Shrink modifies wavelet coefficients with soft threshold function. This method reduce the quality of images due to a constant bias between the original wavelet coefficients and modified wavelet coefficients. Considering that the amplitude of noise coefficients reduced with the increase amplitude of wavelet coefficients,we proposed an adaptive threshold function bearing scalability. Our method overcome the constant bias produced by soft threshold function and can provide the adaptive bias between the original wavelet coefficients and modified wavelet coefficients according to the amplitude of wavelet coefficients. Bisection method has been adopt to get the optimal value of the dilation factor m in accordance with the characteristics of images. The experimental results show that the improved threshold function can get better denoising effect than classical soft threshold function.