基于样本噪声响应法的轮廓波消噪算法

Sample noise response method based contourlet transform denoising algorithm

分析了轮廓波消噪的有关性质,提出了适用于轮廓波消噪确定子带阈值收敛因子的样本噪声响应法。该方法根据样本噪声(标准高斯白噪声或者特定冲激信号)作用在每个子带上的统计特性,得到每个子带的收敛因子,使用该收敛因子对3σ(或4σ)准则进行修正来确定不同尺度不同方向子带的硬阈值门限。图像消噪实验结果表明:无论在峰值信噪比方面还是在视觉效果方面,本方法均可以取得比较满意的消噪效果;对于尺度较大的图像,可以极大地加快消噪速度并减小内存需求;硬阈值消噪之后使用自适应维纳滤波,峰值信噪比会有一定程度的提高。

Some characters about contourlet denoising are analyzed. A sample noise response method suit able for determining the subband threshold factors of contourlet transform denoising is proposed. Using this method, the shrinkage factor of each subband can be acquired according to every subband statistical character driven by the standard Gaussian white noise or certain impulse signal. The hard thresholds of every directional sub-band of various scales are determined by modifying the 3a（or 4a） rule in terms of the corresponding effect factor. Image denoising experimental results show that using the denoising proposed method, the denoising re- suits including the peak signal-noise ratio （PSNR） and the quality of visual effect are perfectly satisfied. The de- noising speed can be accelerated and the memory for the program is reduced especially for huge images. Casca- ding the adaptive Wiener filter with the hard threshold denoising can somewhat improve PSNR.