边界保存的二进小波图像去噪算法

Edge-preserved image denoise algorithm based on dyadic wavelet transform

根据小波变换系数与图像边界的关系,提出了一种基于二进小波变换的图像去噪算法。首先用二进小波在不同尺度上分解图像,在低频部分上提取图像边界,根据图像边界与小波系数的关系,估计对应尺度上高频部分的噪声的方差,用Oracle估计子估计图像的二进小波系数,用估计出的二进小波系数重建图像。实验表明,该算法能够有效地去除各种分布的图像噪声。

A new algorithm of image noise reduction based on dyadic wavelet transform was proposed. Firstly, extract edge from the smooth part of image at different scales. Then estimated the variance of noise in every subband, computed the Oracle estimator and estimate dyadic wavelet coefficients of image at the scale. Finally, restored the image from the estimated coefficients. Experiments show this algorithm can eliminate noise with different distribution.