Research Of Image Inpainting Method Based On Criminisi Algorithm

Digital image inpainting refers to the prediction and reconstruction of the damaged area information with the help of the information of the known region of the image,restore the original image integrity and visual effects.At present,there are two main methods of digital image inpainting:image inpainting based on partial differential equation and image inpainting based on texture synthesis.In the former,the known region information of the image is spread to the unknown area in pixels,It can obtain the ideal effect in the repair of the damaged image with small scale,but it is easy to produce the fuzzy effect when the structure is more complex and the damaged area is larger.The image inpainting method based on texture synthesis usually takes the image block as the unit,and uses the known image information to synthesize the unknown information,which can repair the image structure information and texture information at the same time,so it has a good effect in repairing the image whose damaged area is langer.However,there are some defects in the Criminisi algorithm,such as the unreasonable repair sequence and the inaccurate matching rules and so on.In this paper,by analyzing the shortcomings and deficiencies of the Criminisi algorithm,the following improvements are proposed:(1)In order to overcome the shortcomings of the repair sequence,the priority function is defined in terms of segments,and the structure priority line is introduced;(2)Aiming at the shortage of matching method,this paper presents the matching algorithm based on multi feature which using color and structure similarity to improve the precision of matching,and introduce gradient weights and boundary distance weights restriction factor to select Matching block which can better meet the requirements of human vision;(3)In order to solve the problem of the size of the matching block template,this paper puts forward a method to adjust the size of the template window adaptively according to the complexity of the local texture and structure information of the image in the repair process;(4)The block effect elimination method based on gray fusion is proposed to deal with the block effect caused by matching error.