Research On Digital Image Inpainting Algorithm

Digital image inpainting techniques have been applied widely in many areas withthe development of digital image technology, such as image compression, photorestoration, video effects, heritage, etc. Digital image inpainting techniques areaim to fill the damaged areas of the image and keep the image visual consistencyand integrity. The standard of the image after inpainting is natural, integrity andit is difficult for human eyes to find restored marks.Chapter1of this thesis briefly introduces the research background，significance，status of image inpainting; and chapter2introduces some classical methods for imageinpainting.Chapter3first gives a detailed derivation of each step for the TV model; then givesan improved image TV model for solving the problem of more iterations and lowerspeed of TV model. This improved model increases the number of known pixel in theprocess of calculating points to be repaired, which takes advantage of knowninformation fully and decreases the iterations and improves the repair efficiency.Chapter4presents an improved adaptive priority sample block matching algorithmby adding information item for avoiding the problem of the deviation accumulation andsame sample block size of Criminisi algorithm, which makes the repair order morereasonable. In the process of selecting sample block, this algorithm determinesautomatically the size of the sample block repaired through the differences of the localstructure of the block to make the repair process more reasonable. The improvedalgorithm not only improves the repair effect greatly, but also improves the repairefficiency significantly.Chapter5gives a summary of this thesis and some thinking and outlook of thefuture research of image impainting algorithm.