The Image Structure To Maintain The Variational Pde Processing Model And Applied Research

With the research of image processing tasks such as image restoration and image denoising, the research of image models based on variational PDE methods has been draw a great attention from international views. How to preserve the original structures of an image during the image processing is one of the most important issues and it has become a popular topic in image processing.With the applications for image restoration, the paper focuses on the image processing which based on Variational PDE models with structure preserving ability and has an overview of the research status of variational PDE models and algorithm. Based on the idea of local and contextrual structure drived PDE model and nonlocal model, the paper includes the image restoration algorithm based on local and contextrual structure drived PDE model, the structure preserved TV model and its adaptive fidelity term and the nonlocal regularization of image restoration model. The following results are achieved:In the aspect of local and contextual structure drived PDE model, the classical linear and nonlinear diffusion PDE models are invesitigated firstly and the filtering mechanism and denoising performanace are discussed. Based on the discontinuities of image structure, two distinct measures according to the local structure and contextual structure are introduced to establish the bilateral filtering mechanism. Furthermore an iterative algorithm for adaptive smoothing is derived. A lot of experimental results show that the local and contextural filtering mechanism can provide better structure preserving ability in image denoising.In structure preserving Total Variation model and improved TV model with adaptive fidelity term, the structure preserving adaptive fidelity term is discussed based on the classical TV model. Then a numerical algorithm is presented for our experiment. The results prove that in the respect of structure preserving, the improved model has a better performance of image texture preserving.For the nonlocal regularization of image restoration model, first the nonlocal mean filter is introduced. Secnd with the idea of nonlocal filtering we propose two kinds of regularizing functionals, derive their Euler-Lagrange equation and analyse properties of the linear operator L. At last the similarity function based on pixel Euclidean distance, pixel intensity and both Euclidean distance and intensity are proposed as weight function in regularizing functionals. After comprehensive compare, our experimental results prove that with different weight functions, the improved regularizing functional is effective for images with different textures and details.