Image Inpainting Algorithms Based On Partial Differential Equation

Image inpainting is an important content, but also the present computer,applied mathematics, electronic information related field one of the keyresearch contents. Image in process is according to the image of the knownarea of relevant information, together with an inpainting algorithm forrecovering the inpainting of the area of the information, to complete thewhole image of "reasonable", satisfying visual effect requirements.According to image inpainting theory and image inpainting domaindifferent feature information, image inpainting algorithm can be roughlydivided into two kinds: one is based on the structure of information imageinpainting algorithm, another kind is based on texture synthesis imageinpainting algorithm. With the increasingly digital life products andelectronic computer of booming development, image inpainting techniquesa wide range of applications, mainly used in military and defense ofmodern industrial production, medical image processing, photography,satellite remote sensing telemetry, the criminal investigation, gamedesign and production of special effects in movies and other fields, andhas become the focus of attention. This area of research, foreign has beenbooming development, domestic start later.This paper first introduces digital image inpainting technology researchbackground and research significance, given the image inpainting andinpainting the mathematical description of the image of the basicprinciples, the best guess principle, bayesian framework theory, theimage of the inpainting algorithm evaluation standards and thevariational method and functional extreme value and related basicknowledge. And then based on the partial differential equations of the image inpainting algorithm, and discusses several model based on partialdifferential equations of the inpainting algorithm, including BSCB model,TV model, the CDD model for three inpainting model detailed theoreticalanalysis, numerical implementation process and the derivation ofsimulation experiment, summarizes the advantages of various algorithmsand their respective deficiency, and in the light of the TV model has"stept effect" put forward a kind of improved the inpainting algorithm,in combination with the isotropic and anisotropic diffusion ofinformation communication manner, the hierarchical image inpainting, useof regional frequency difference in different area realized usingdifferent iterative equation, thus effectively avoid the original TValgorithm introduces ladder effect, to improve the iterative efficiency.The experimental results show that this algorithm and the TV modelalgorithm, which has the same in inpainting effect, under the premise ofthe improved algorithm avoid stept effect and has better than TV modelof the inpainting of the speed.