Study Of Image Inpainting Based On Learning

In this paper, we construct a actual system of image inpainting based on the image inpainting system model[1] which was proposed before, in order to repair more types, more broken images in the different field and restore them more efficiently ,because there is no one “universal” or more general algorithm can repair all types of images. The proposed system integrates the commonly used several typical algorithms, when using these algorithms to repair the image, adding some knowledge of human-assisted analysis to help users select the appropriate algorithm for image restoration. It is necessary that the broken images needs classifies subjectively and objectively before inpainting and the result of image inpainting needs estimate subjectively and objectively after inpainting. The estimation results of image repairing can analyze the advantages and disadvantages of the various algorithms; it can provide users with appropriate algorithms selected. This system also has the self-learning function to make the system more flexibility and practicability.