Research On The Applications Of Sparse Representation Theory In Medical Image Processing And Analysis

In recent years, one of the latest developments in the signal processing is the sparse representation theory. Sparse representation theory has been widely used in image process-ing, pattern recognition, computer vision and so on. At the same time, sparse representa-tion theory based sparse coding and sparse regularization have been growing emphasis on the majority of researchers in medical image processing and analysis. Using sparse the-ory technology can solve the traditional medical image denoise, enhancement and analysis problems.Medical image processing and analysis mainly includes image restoration, enhance-ment, segmentation, image sequence analysis and other issues. In this paper, several prob-lems in medical image processing and analysis are studied based on the sparse representation theory. Medical image restoration, image enhancement and image sequence analysis were discussed, specific results are as follows.This paper solves ultrasound medical image denoising problem based on speckle noise model and image sparse coding. The paper uses the (?)2sparse coding representation which is sparser than traditional (?)1to describe the image data. In Bayesian-MAP framework, based on the statistical distribution of the coding, appropriate threshold shrinkage algorithm is designed to suppress noise and recover image. Experimental results show the algorithm can effectively recover medical ultrasound image, suppress the speckle noise, enhance the edge and keep the detail.This paper studies the bias field adaptive correction in medical imaging enhancement problem. To solve the problem for the determination of bias field model parameters in the traditional correction methods, the link between the bias field and the observed image of medical imaging is analyzed. A reasonable assumption is made to achieve the image enhancement. Sparse constraint adaptive correction and enhancement model is built based on the bias field imaging. The enhancement model uses two constraints to describe the bias field and the ideal image. The ideal image is required to be piecewise constant, and the bias field is required the gradient smoothness. Iteratively evolutes observation image surface to approximate the bias field adaptively to avoid the determination of bias field model parameters. An alternative optimize method is used to search the optimal to achieve correction enhancement purpose. Simulation and real data experiments show the algorithm can achieve medical image enhancement better.This paper studies the medical image sequences automatic analysis technology. Dis-cussing traditional medical image sequences factor analysis methods, a sparse nonnegative matrix factorization based image sequences factor analysis method is presented in this pa-per. The method is improved to solve problems in traditional factor analysis methods, which are the solution of nonnegative and uniqueness of the solution. Ensuring the premise of non-negative result, a sparse regularization constraint term is proposed not only to describe the physical structure of independence, but also can limit the structure of the solution. And then a new factor analysis model can be constructed. Using sparse nonnegative matrix factorization algorithm to solve numerical calculation. Simulation results demonstrate the effectiveness and stability of the algorithm. The results are greatly improved than traditional methods.Then the new method is the further applied to the liver ultrasound perfusion imag-ing data, and the results have been recognized by the doctor’s diagnosis, which verify the feasibility and accuracy of the technology.At last, we summarize the presented work. And we analyze the imperfect parts, discuss the future work.