Research On Image Inpainting Technology Of Large Area Damaged

With the development of image processing related technology in the field of computer science,digital image inpainting has become one of the branches and has gradually become the focus of research in recent years.In daily life,it is very easy to encounter image damage problems,such as the destruction of precious art paintings,old photographs creases and stains,the removal of redundant objects and post-film production.It will produce a large area of damaged areas,then need for image inpainting technology makes the human can not perceive the image has been changed.Therefore,large area damaged image inpainting has important research value and significance.The Criminisi algorithm is a typical example of repairing large area damaged image.Its important ideas and simple design are important documents of Chinese and foreign scholars in recent years.Although this method achieves a breakthrough in image inpainting of large area damaged,the effect of the method for the large area of the straight line structure and texture coexistence of damaged is poor.The current improvements are made to ensure the correct connection of the structure within the damaged area,but can not reasonably repair the texture part.In order to solve the above problems,this thesis based on the Criminisi algorithm to make the following improvements:Firstly,the local variance of luminance is introduced into the priority,which ensures the order of repair in accordance with that of manual repair,so as to achieve more accurate effect.Secondly,on the selection of the optimal matching blocks,the automatic adaptive choosing of template sizes is designed to avoid the repairing quality problems resulting from the over or under size of blocks,then further makes the image inpainting more delicate.A large number of simulation experiments show that the part of the image structure after restoration is basically correct,and the part of the texture is also very reasonable,which is consistent with the human visual perception.Finally,aiming at the authenticity and fairness of the evaluation of the image quality,an evaluation method of subjective and objective combination is used in this thesis.Both from the objective point of view and the subjective point of view,experimental results show that the improved algorithm is of higher quality and better effect.This method effectively solves the problem that part of texture is not restored reasonably.