Paper Info
Title
Progressive completion of a panoramic image
Abstract
Abstract This paper proposes an exemplar-based modified priority belief propagation (MP-BP) method to complete a stitched panorama. The result is an image with a rectangular boundary in which the missing area is filled by a visually plausible background that copies the appearance of the source region. Global optimization is usually preferable to greedy algorithms for image completion. Komodakis (IEEE Trans. Image Process. 16 (11): 2649–2661, 2007) proposed a priority BP method, which is expressed in the form of a discrete global optimization problem with an MRF energy function. However, this priority BP method cannot be directly applied to stitched panoramas completing problem because there are very few useful messages. Instead, a progressive way to expand the boundary is proposed. The main contributions of this study are: (1) it defines a progressive way in which to apply a Markov Random Field (MRF) model to complete panoramas; (2) it defines a priority term that integrates the concepts of “confidence” and “breadth first”, to guide the filling order; and (3) a restricted source region and clustered candidate patches are used, to alleviate computation complexity. A number of examples of real stitched panoramas demonstrate the effectiveness of this algorithm. The results compare favorably with those obtained using existing techniques.
Year
DOI
Venue
2017
10.1007/s11042-015-3157-5
Multimedia Tools and Applications
Keywords
Field
DocType
Panorama,Image completion,Belief propagation (BP),MARKOV Random Field (MRF)
Computer vision,Global optimization,Computer science,Markov random field,Panorama,Breadth-first search,Greedy algorithm,Artificial intelligence,Computation complexity,Belief propagation,Global optimization problem
Journal
Volume
Issue
ISSN
76
9
1573-7721
Citations 
PageRank 
References 
1
0.37
19
Authors
3
Name
Order
Citations
PageRank
Shwu-huey Yen1429.07
Hao-Yu Yeh210.37
Hsiao-wei Chang331.78