Research On Exemplar Block-based Image Inpainting Algorithm

Image inpainting is a hotspot in humans working and daily life. Image inpaintingis one of the important parts in image processing, which refers to reconstruct thedamaged image or remove the unnecessary part from the whole image. This techniquehas found widespread use in applications such as cultural relics protection,multimedia editing, video privacy protection, and unnecessary objects removing.Image inpainting methods are mainly divided into two categories: image inpaintingbased on non-texture structure and image inpainting based on texture synthesis.Exemplar block-based image inpainting algorithm is a typical texture synthesisalgorithm based on image inpainting techniques.Criminisi has proposed an exemplar-based image inpainting algorithm. It usesthe method of texture synthesis to remove the obstacle in the image, and reaching agood result. Criminisi algorithm mainly considers inpainting order in the target area.When we are filling the target area, we firstly calculate all the priority of theinpainting block on the boundary of the damaged area, and select the block with thebiggest priority for early repairing. Criminisi algorithm has a simple and fastexperimental process, and a better result. But in the practical application, theestimation of image edge is affected by the high-frequency information of the imagetexture, which leading to an error inpainting order, and affecting the final result.In this paper, we improve the priority computation formula, by leading an αfactor to enhance the proportion of data. Therefore, the algorithm is more sensitive tothe detail of image texture during the inpainting procedure. We use addition instead ofmultiplication to prevent rapid decay, which gives us more reasonable inpaintingorder，so it can enhance the inpainted effects.On the base of theoretical analysis, we carry on a full of experiment about theproposed improvements. Experimental results show that the proposed algorithm canbe satisfied by the inpainted effects.