The Research Of Sparse Decomposition In The Field Of Image Denoising

In the process of obtaining the image, which will be mixed with some noise that will damage the quality of the image. So, image noise must be processed. But the traditional image denoising methods are based on assumption of low frequency image and high frequency noise, but this assumption is not always true. In order to find a better denoising method, the sparse decomposition applied to the image denoising. By choice a better coherent ratio threshold achieved image denoising.Firstly, Over-complete atomic library select Gabor atom to construct, then using the improved FFT-MP algorithm for image decomposition, selecting coherent ratio threshold method as the termination of iterative conditions for sparse decomposition; Secondly, the residual component get in the process of decomposition as noise be discard; Finally, the image is reconstructed by a linear combination of the best match atom. Experimental results show that the algorithm in this paper is not limited by the noise level, and superior to traditional denoising algorithm in the visual effects.Research in this paper include the following aspects:1. Researches and improvements of traditional denoising algorithm. By comparing the experimental results of mean filter median filter wiener filter, analyzing the pros and cons of three denoising algorithm, and lead to a hybrid filter.2. Improved FFT-MP algorithm. for MP algorithm has a slow speed, the paper use a cross-correlation replace inner calculation, and then implementation by the fast Fourier transform, to find the best atoms. Simulation results show that the algorithm is faster than second MP algorithm, and with a few atoms can exhibit most of the characteristics of the image.3. This method applied to simulation experiments, compare and analyze the algorithm results, and find out the cause of the results.