Paper Info
Title
Screen Window Propagating for Image Inpainting.
Abstract
Image repair often has errors in filling. An exemplar-based technique has the advantage of filling texture and structure simultaneously, but there exist some problems with this, such as not considering the overall situation of inpainting, the unreasonable calculation of priority, and random selection when there are multiple candidate patches. In view of these situations, this paper proposes a method of screen window propagating where the upper layer guides the lower layer inpainting from the top layer by means of multiple resolution decomposition. Screen window propagating aims to constrain the inpainting of the lower layer and maintain the overall profile. Improving the priority calculation using the upper repair result can lead to getting a better repair order. When there is more than one candidate patch, the structural similarity index measure (SSIM) is used to obtain the best candidate patch. We consider four cases of rotation and inversion when looking for the best exemplar. The experiments show that our method can obtain more satisfactory results than other methods based on subjective and objective indicators, such as peak signal-tonoise ratio (PSNR), SSIM, and feature similarity index measurement. Screen window propagating for image inpainting aims toward image restoration. Whole contours are preserved and the lower layer can utilize more upper layer information. Our method considers information from the whole and its parts.
Year
DOI
Venue
2018
10.1109/ACCESS.2018.2876161
IEEE ACCESS
Keywords
Field
DocType
Exemplar-based technique,pyramid decomposition,guidance strategy of screen window propagating,exemplar matching
Computer vision,Microsoft Windows,Computer science,Inversion (meteorology),Inpainting,Sampling (statistics),Artificial intelligence,Image restoration,Image resolution,Maintenance engineering,Distributed computing
Journal
Volume
ISSN
Citations 
6
2169-3536
0
PageRank 
References 
Authors
0.34
0
6
Name
Order
Citations
PageRank
Liu Huaming132.77
Bi Xuehui221.73
Guanming Lu3299.43
Weilan Wang4911.75
Jingjie Yan543.14
Zhengyan Zhang603.38