Research On Image Inpainting Based On Texture Synthesis Method And Partial Differential Combination

Image restoration in image processing technology is a hot topic, has important applications in many areas, in the field of future research still has important research value. This text studies the algorithm of structural and texture type in the algorithms of image restoration, narrates detailed and analyzes their advantages and disadvantages, mainly to do the following research areas:First, the text descripts the overall variational method (TV model) detailed, analyses the model for this kind of repair, after the overall improvement of the variational method to find a model that requires the use of improved partial differential equations. In the course of the proof of partial differential equations, equations of defects found in some areas, and take a two-dimensional interpolation to compensate for deficiencies in order to determine the method to repair damaged image in the actual practicality.Secondly, the general repair algorithm based on texture synthesis made further improvements. Restoration algorithm for the pixel block proposed by Criminsi can’t be determined corresponding to the size of the template matching, as well as to maintain the boundary can’t be achieved in line with the sampling region, presents a texture synthesis algorithm, I taking an adaptive block matching method selected, The adaptive factor is introduced into the gradient information in the image which determines the need to match the size of the selected block.Finally, the text decomposes the use of image decomposition ideas to fix an image into two parts, namely the structure of part and the part of the texture map. TV model algorithm uses an improved view of the portion of the structure to repair and maintain their sharpness of the image boundary; then taken to repair the block matching algorithm based on adaptive selection of textures repair parts, parts of the image to preserve the texture, the results of the above two algorithms obtained were superimposed to obtain the final results of the recovery. Experimental results show that the use of image decomposition method is obviously better than a single algorithm.