Segment And Inpaint An Image With Partial Differential Equations

Partial Differential Equations (PDE) are extended used many physical modeling problems such as particle transportation, gas diffusion and liquid flow. From around 1985, PDEs image processing and image analysis became an active topic in related research area. Typical applications of PDEs in image processing are image denoising, image segmentation, 3-dimentional image redering, and so on.This work traced the development in PDE's image processing, stressed on PDE's image segmentation and image inpainting, and implemented several classical algorithms in this trend. The essence of PDE's segmentation is to drive a curve to the expected edges via various of strategies of curve evolution. This paper analyzed Mumford-Shah piecewise constant model and several active counter models. PDE's inpainting models drives image information along isotropice lines. Therefore, our model segments an image into structure subspace and texture subspace, then inpaint the structure with BSBC-inpainting algorithm and texture simulation in texture subspace. Another algorithms is non-PDE's EBI-inpainting algorithm.This paper summarized some skills of image PDE's calculation within levelset framework, and exhibited our results of image inpainting. We warned of some possible problems of inpainting's application to medical images.