基于去频谱混叠Contourlet变换的层内局部相关性图像降噪

A novel intra-scale local correlation image denoising method based on non-aliasing contourlet transform

提出了一种基于去频谱混叠Contourlet变换的层内局部相关性图像降噪新方法。含噪图像经抗混Contourlet多尺度变换,得到一个低频逼近子图和一系列不同尺度、不同方向的高频细节子图,充分利用变换域同层同方向子带内信号系数相关性强、噪声系数无相关性的特点,采用强局部化零均值高斯分布模型对高频细节子图进行降噪处理。实验结果表明,该方法计算效率高,能克服Contourlet变换中的频谱混叠,避免了重构图像出现"划痕"现象。无论是PSNR指标,还是在视觉效果上,该方法的去噪性能均好于Contourlet去噪、Contourlet域HMT去噪和基于抗混叠Contourlet变换的硬阈值去噪,在有效去噪的同时,具有很好的图像边缘和细节保护能力。

A novel image denoising method based on non-aliasing Contourlet transform is presented by using zero mean Gaussian model. A noisy image is decomposed into a low frequency approximation sub-image and a series of hign frequency detail sub-images at different scale and direction via multi-scale non-aliasing Contourlet transform. In the transform domain, the intra-scale correlation in same directional sub-image of the signal coefficients is strong, and there is no intra-scale correlation for noise coefficients, so the noise in the high frequency detail sub-images is removed by using a strong local zero mean Gaussian distribution model. Experimental results show that the proposed scheme has higher operational efficiency, and it can overcome the aliasing in Contourlet transform and avoid ＂scratching＂ phenomenon in the reconstructed image. Whether PSNR index or in visual effect, the proposed scheme outperforms the traditional Contourlet transform denoising, Contourlet domain HMT denoising and the hard threshold denoising excellent balance between suppressing noise effectively and based on no-aliasing Contourlet transform, and can achieve an preserving as many image details and edges as possible.