Study Of The Image Denoising Based On HVS And TV Models

As an objective reflection to the natural scenery, image has been more and more important for people to know the whole world and themselves. In fact, noise not only makes the image bad but also impacts the following image processing. Both the traditional analog images and the digital images are all subjected to different kinds of noise pollution inevitably. As one of the image pre-processing technologies, image denoising plays a very important role. The quality of the denoised image has an influence on the whole image processing. Based on this, emphasis study is given on the theories of HVS and PDE used in image denoising. From this, an improved image denoising method is presented as well as the investigation on the development and combination between the HVS and image estimation. The main work and originality in the thesis are listed as follows:1. Noise classifications and models are summarized. The image denoising theories and algorithms are summed up and validated in detail, which are sorted in two kinds: linear and nonlinear.2. Emphasis investigation and discussion are carried out on features of HVS used in image denoising. The noisy sensitive coefficient which is base on HVS is calculated through Fourth-order Cumulant, and an improved traditional image denoising method is presented which has been proved more effective in image denoising and can keep the details better. Compared with the traditional linear and nonlinear algorithms, the improved method has a better filtering and vision effect.3. Combined with theories of PDE,detailed research is given on the development and applications of PDE in image denoising. The PDE method based on TV models is applied in salt and pepper noise denoising. Experimental results prove that this method can get rid of the salt and pepper noise and achieve a better compromise between detail-pretection and denoising.4. Detailed summaries on traditional image estimation methods are given as well as research on the combination between subjective and objective estimation methods. Based on the analysis and investigation to the image estimation method combined with HVS, several methods which do well in image estimation are given. A series of experimental comparisons and performance are carried out to the improved algorithms in this thesis, which can prove the image estimation method combined with HVS is rational.