Rotation Invariant Common Sketch Extraction
- Resource Type
- Conference
- Authors
- Li, Si-Yuan; Yang, Ya-Li; Sun, Yun-Feng; Hui, Dan
- Source
- 2015 International Conference on Computer Science and Applications (CSA) Computer Science and its Applications, International Symposium on Computer Science and Applications (CSA), 2015 International Conference on. :82-85 Nov, 2015
- Subject
- Computing and Processing
Companies
Electronic mail
Shape
Mathematical model
Image retrieval
Correlation
Refining
common sketch extraction
rotation invariant
self-similar
- Language
In this paper, we present a new method to sketch the common among several images. Our method captures rotation invariance by extending the local self descriptor to rotation invariance and proposes a easy method to detect a roughly similar region across the images. Our method is composed of three stages: (i) Detecting a similar region which the proportion of the common is as large as possible across the images. (ii) Estimating the rotation angle by matching the descriptors across the similar region, then aligning the images. (iii) Generating a clean and compact binary sketch. The experiments demonstrate that our method can handle the rotation between common and obtain a clear common sketch.