基于压缩感知的图像降噪处理研究

Image Denoising Method Based on Compressed Sensing Theory

在农产品图像的动态采集中,可能会出现粘结、叠加及背景干扰等一系列缺陷,同时稀疏采样的图像也可能是不完整的。对于这个棘手的问题,由压缩感知理论可以找到答案。压缩感知理论首先对采集的图像进行稀疏表达,然后选取适合图像的最优小波基,采用凸优化理论及其算法,可以得到花生图像的特征点(降噪点)并进行处理,从而完成噪声的去除。为此,在压缩感知理论的基础上,提出了运用快速迭代阈值收缩(FISTA)算法进行去噪处理,与其他的图像降噪方法相比,体现了速度快、效率高、去噪效果好等优势。

In the Agricultural images obtained by the dynamic acquisition, there exist many defects such as bonding, su-perposition and the background interference.In addition, the image obtained from the sparse sampling of the peanut is not complete.All above problems can be solved by the compressed sensing theory.First, compressed sensing theory have a sparse expression for the capture picture, and then select the best wavelet basis for peanuts image, the use of convex optimization theory and algorithms can get the feature points ( noise points) of peanut image.Finally, the image noise can be removed completely .Experiments show that the compressed sensing theory in the process of peanut image denoising possesses many excellent properties such as high efficiency, better denoising result and other advantages com-pared with the traditional methods of image denoising.