The Research Of MRC Image Denoising Based On Prior Learning Of Similar Block

With the development of the Cryo-electron microscope technology, the demand for the imaging precision of the biological macromolecule is became higher and higher. Under the condition of now, the projection image of the biological macromolecule which is collected by the Cryo-EM technology that the contrast is low, the signal to noise is low, image blurring,and not easy to distinguish single particle from background, The corresponding image processing technology is lagging behind, which restricts the development of electron microscopy technology. At present, the Cryo-EM image denoising is mainly for single particle image statistical average in the Fourier space. But this method does not take into account the noise mixed in the original Cryo-EM image which will affect the extraction of single particle. In additional, the statistical average cannot effectively eliminate the noise. Therefore, it is necessary for the original Cryo-EM image denoising.The difficulties of image denoising mainly includes the following two aspects: firstly, it is difficult to distinguish the noise and signal effectively. Secondly, there is no well-known image denoising model. The collected data by Cryo-EM technology contains many microscopy images, the single particle images which were used for three-dimensional reconstruction were extracted from these original images. so, MRC image denoising can improve the quality of the image, for the subsequent particle extraction, the two-dimensional image classes and average provide highly quality image, which contribute to get a high resolution three-dimensional structure. MRC image denoising has important significance for three-dimensional reconstruction.Innovations of this paper are as follows:1. For the limitations of using Euclidean distance and Mahalanobis distance to measure similar blocks, In this paper, a method of image block similarity measurement based on geodesic distance is proposed. By calculating the geodesic distance between the two image blocks to determine whether the two image blocks are similar, the proposed method can accurately search the all similar blocks of the referenced block in the entire image domain.2. For the insufficient of BM3 D algorithm that perform collaborative filter for every block group in transform domain, This paper proposed a method based on similar block prior learning. Firstly, Through prior learning get K Gaussian component, Then, using bayesian maximum a posteriori probability for a similar block group to choose a suitable Gaussian component,the covariance matrix of the Gaussian component is decomposed by SVD, and the dictionary of similar block group is obtained. Weighted sparse coding model is used to solve the sparse coding of similar block groups, using the sparse representation of the natural image denoising the similar blocks. The proposed method can avoid limitation of image sparse representation using a global dictionary. In this paper,using proposed denosing method for every similar block group separately. When a number of estimated values appear in a position, the final estimate of the pixel is obtained by using the weighted average. The experimental results show that the proposed denoising method in the paper can be used to the MRC image denoising, and the PSNR value can reach 45.65, the proposed method is better than other denoising algorithm,using denoised Cryo-EM image for three-dimensional reconstruction, we can get the three-dimensional structure of biological macromolecules.