Light In Case Of Image Restoration Techniques

Nowadays,with the fast development of Digital technology,People have more and more interest in photo-taking and post-processing of Images.As a new and effective function,image completion would become the necessary plug-in of Image processing software in the future.However,current image completion skills cannot be used to deal with some special situations:for example,not able to predict the illumination or depth information of the unknown area.This paper proposes a new framework for image completion:Decompose the image into two images based on Intrinsic image model:shading image,which describes the illumination and shape of the surface,and reflectance image,which describes the reflectance of the surface.The images with only one intrinsic characteristic can be dealt with more easily.After processing the two intrinsic images,we can combine them back to the result image.This paper applied this framework to images under two conditions:painting or text removal from the surface of the object under illumination;object removal under illumination.While dealing with intrinsic images, I used current image completion algorithms,including structure propagation by global optimization and texture propagation by graph-cut patch-based texture synthesis algorithm.Experimental results show that it's an effective framework of image completion with high efficiency.However,our algorithm is based on the result of intrinsic image model.Since recovering intrinsic images is a new research area and the results are limited in some ways,the result of our algorithm may be impacted a little,too.