A Description Method for MSER with SIFT Descriptor

Abstract:

Maximally Stable Extremal Regions are robust to complex affine distortion and illumination changes between reference image and real-time image. On the basis of deeply research on the SIFT descriptor, this paper propose a description algorithm for MSER using SIFT descriptor. The Second central moment is used in the algorithm to make ellipse adjustment for each irregular MSER. Then a self-adaptable rectangle area, whose side is proportional to the minor axis of the ellipse, is constructed encircling each ellipse centre. Finally, a SIFT feature vector is formed to express the MSER, after processing the statistic characteristic of all pixels’ gray gradient in the rectangle aero. A reasonable range of a self-adaptable rectangle area side is presented in the paper. Matching experiment results show that our algorithm is highly distinctive and stable.

• Periodical: Applied Mechanics and Materials (Volume 127)
• Pages: 115-120
• DOI: https://doi.org/10.4028/www.scientific.net/AMM.127.115
• Online since: October 2011
• Authors: Xiao Hu Fan, Wei Ting Lin, Juan Cao, Ben Ling Li, Yan Si
• Keywords: Affine Invariant, Descriptor, Maximally Stable Extremal Regions, SIFT

References

[1] O. Chum,J. Matas,M. Urban T. Pajdla (2002). RobustWide Baseline Stereo from Maximally Stable Extremal Regions. In British Machine Vision Conference, pp.384-396.

DOI: 10.5244/c.16.36

[2] K. Mikolajczyk, Schmid C. An affine invariant interest point detector. In Proceedings of the 7th European Conference on Computer Vision. Copenhagen, Denmark, (2002).

DOI: 10.1007/3-540-47969-4_9

[3] Cordelia Schmid, Krystian Mikolajczyk. Scale & Affine Invariant Interest Point Detectors [J]. International Journal of Computer Vision. 2004. Vol. 60(01): 24.

DOI: 10.1023/b:visi.0000027790.02288.f2

[4] Per-Erik Forss´en, David G. Lowe. Shape Descriptors for Maximally Stable Extremal Regions. BMVC2002: 113-122.

[5] Štěpán ObdrŽálek, Jiřĺ Matas. Object Recognition using Local Affine Frames on Distinguished Regions. David G. Lowe. Distinctive Image Features from Scale-Invariant Keypoints[J]. International Journal of Computer Vision. 2004. Vol. 60(2)(Kluwer Academic Publishers): 20.

DOI: 10.1023/b:visi.0000029664.99615.94

[6] Wu Yufeng. The application of Maximal Stable Extremal Regions in image searches[D]. Da Lian. Da Lian polytechnic university. (2008).

[7] O. Chum,J. Matas,M. Urban T. Pajdla (2002). RobustWide Baseline Stereo from Maximally Stable Extremal Regions. In British Machine Vision Conference, pp.384-396.

DOI: 10.5244/c.16.36

[8] Mohan Trive, Erik Murphy-Chutorian. N-tree Disjoint-Set Forests for Maximally Stable Extremal Regions. the British Machine Vision Conference. Edinburgh, 2006: 278-286.

DOI: 10.5244/c.20.76