Research On Image Denoising Based On Morphological Component Analysis

Digital image denoising, is one of the most basic topic in Image Processing. The difficulty is there is no common method for all kinds of images, if performance is given priority. The traditional denoising approach is adopting denoising algorithms regarding the feature of the image and noise. If the feature is unknown before the algorithm selection, the image denoising is just done by trying algorithms. This dissertation is proposing a novel algorithm, based on morphological component analysis (MCA) theory, which possesses good denoising performance and widely applied range.Image is usually composed of various components, most of which contain cartoon component and texture component. The target of blind source separation (BSS) is to separate these components from each other. During the process, if the image noise, considered as a special component, can be separated and thrown away. As for the image noises among other separated components, certain denoising algorithms will be applied according to the feature of the component. The last step is two reconstruct the image by combining all separated components together. Through lots of experiments on this algorithm, by contrast with the traditional ones, the performance of denoising and applied range is both better.At first, this dissertation introduces the basic theory of image denoising and traditional algorithm. Through certain experiments and algorithm analysis, the pros and cons of traditional algorithms are concluded. Then, Blind Source Separation is briefed through its basic theory and applications. The key point is MCA theory and algorithm, which are validated by experiments on one dimension and two dimension signals. Based on what is discussed, a novel algorithm is proposed, on which different type of images with different type of noise is experimented. From the experimental results, the performance, complexity, and applied range of the algorithm is analyzed. At last, certain conclusion is proposed and certain prospect of unfinished work and functionality is made.