.png){kind=icon}
.png){kind=icon}
Return
Data-driven image completion for complex objects
.png){kind=icon}
PRE
AI
DOI：10.1016/j.image.2017.04.011.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Image completion is challenging for objects with complex structure and texture as the missing information is unpredictable. In order to provide a feasible solution for such situations, we introduce a new system of image completion using a data-driven approach with simple user guidance. Our system reconstructs the degraded image region by employing a coarse-to-fine object and patch matching process. At the object matching stage, the system retrieves the object information not self-contained in the degraded image by searching for the best matching object via the user's specification of the missing contour. Then at the patch matching stage, the degraded region will be repaired by similarity-constrained contour propagation and bidirectional region filling. Extensive experiments show that our system outperforms state-of-the-art image completion methods in terms of repairing object images. Moreover, our system is more flexible as it can yield different appearances of the object according to different shapes specified. (C) 2017 Elsevier B.V. All rights reserved.
Keywords:
Image inpainting
Data-driven
Object matching
Exemplar-based
AI Summary
Key information extracted from the uploaded paper, including a brief overview, abstract, background, key highlights, visual analysis, and future outlook.
Journal
S
IF:
2.7
Papers:
2.8K
Citations:
4.2K
Organization
View more organizations