Study On Image Inpainting Method Based On Wavelet Transform Domain

Digital image inpainting is a process of using known information of damaged image to fill reasonably which information of the specific area.The purpose of digital image inpainting is to repair the image while not destroying the integrity images,and recovering the lost information of an image or removing unwanted objects.Now,the digital image inpainting has become a hot topic in computer vision.This thesis analyzes the image characteristics of the wavelet transform domain,and summarizes the image inpainting method based on the classical image inpainting algorithm which using wavelet transform.It takes the research of hierarchical classification method of image inpainting while discussing a range of inpainting models and improvements.The main work and research results as follow:First,in order to overcome the error repair of the boundary of structure and texture in the traditional image inpainting algorithm,the thesis combinates wavelet transform with texture synthesis,and utilizes the characters of the wavelet transform domain coefficients,a novel image inpainting algorithm is proposed.At first,the damaged image is decomposed into low frequency sub-images and high frequency sub-images with different resolutions using the wavelet transformation.Then,the sub-images which reflecting the structural information of the image are reconstructed by fast marching method(FMM)algorithm,while the sub-images which reflecting the texture information of the image are filled-in with texture synthesis by introducing the edge factor based on the characteristics of wavelet coefficient in the sub-images.Compared with the related algorithms,the results of simulation show that proposed algorithm holds good integrated performance.Second,when in the image inpainting process of high-frequency information,the contour curve is easy to be blurring,so the thesis combines the edge structural constraints and characteristics of wavelet coefficients to propose a new image inpainting algorithm.For the high frequency sub-images,proposed algirithm uses the extracted and repaired edges which always contain significant information to constrain the repair of the structures division.Nevertheless,the textures are filled-in with using texture synthesis based on the characteristics of wavelet coefficients of the sub-images.The experimental results show that the hierarchical classification method combined with structural constraints works well for images where the edges were damaged.