基于Shearlet框架的多尺度去噪算法

目前的经典多尺度系统Curverlet、Contourlet存在的主要缺点之一是它们无法将连续性与数字世界进行统一处理,而Shearlet系统是目前多尺度领域内唯一满足这一性质同时还提供对图像的最优稀疏表示的多尺度系统。提出一种用限制频带的Shearlet变换来进行多尺度分析,其主要通过对图像进行快速PPFT变换,以及加权和加窗处理得到Shearlet系数,通过SURE-LET变换进行噪声估计优化分解系数,最后进行Shearlet重构得到去噪图像。实验结果表明,相比于目前的去噪算法,在PSNR、SSIM和时间上,该算法都有一定的优势。

Multiscale denoising algorithm based on Shearlet frame

One of the most common shortcomings of the frameworks of the system Curverlet and Contourlet is lack of providing a unified treatment of the continuum and digital world. Now Shearlet systems are the only systems which satisfy this property, yet still deliver optimally sparse approximations of images. This paper presented a band-limited Shearlet for multi-scale analysis. It used fast PPFT transformation to images, weighted and windowing treatment to get Shearlet coefficient, then optimized the decomposed coefficient of the image noise through the SURE-LET, and finally obtained the denoised image by inverse Shearlet transform. The experimental results show that, compared with the current denoising algorithm, the algorithm takes the certain advantages in the PSNR, SSIM and time.