Research On Improvement Of Criminisi Algorithm Based On Sobel Operator

Most of human information is acquired through the visual nerve system,images become the main media for transmitting information.With the rapid development of information technology and digital processing technology,digital image processing technology emerges as the times require.It has a wide range,including image enhancement and restoration,segmentation,recognition,compression,coding and so on.The main research object of this thesis is digital image inpainting technology,that is,using the pixel information in the known area of the image to repair the damaged area.The image restoration algorithms can be divided into two categories: the first is the image restoration algorithm based on partial differential equation(PDE)for small area damaged region images,and the second is a texture based image restoration algorithm for large size damaged area images.In this thesis,we mainly study the texture synthesis technology based on block matching in the image restoration algorithm.In view of the shortcomings of the existing methods,the improvement is put forward so as to get more conforms to the restoration results of the human vision system.In this thesis,the basic principle of image restoration algorithm based on PDE model is introduced,and the application conditions of this kind of algorithm are analyzed in detail with the simulation results.The effect is very obvious in small scale repair repair,but the large area blurred phenomenon in the repair results can not be overcome when the large size damaged image is repaired,thus verifying the base.The importance of block matching repair technology.Therefore,after introducing the classical Criminisi algorithm in detail,this thesis redefines the priority calculation of the area to be repaired in the Criminisi algorithm by introducing the Sobel operator,and optimizes the repair order,making the repair process starting from the strong edge of the edge,which not only solves the traditional definition of priority.The problem that the image edge and texture identification degree is not high and the calculation results tend to zero suddenly,and effectively reduces the texture extension which is easy to produce in the restoration results.In addition,this paper further improves the matching block search strategy by introducing the standard difference feature of the image and the gradient based structure similarity,thus improving the accuracy of the matching block selection and improving the repair effect.The simulation results show that the proposed algorithm has better image restoration effect than the Criminisi algorithm and its improved algorithms.