Maximally Stable Color Regions Based Natural Scene Recognition
Dongcheng Shi, Guo-qing Yan · 2009
In this paper, we propose a maximally stable color regions (MSCR) based method to perform reliable scene recognition. Firstly, to construct the maximally stable regions, we exploit the MSCR detector for improving the identification of stable regions. Secondly, each detected region is processed properly by using the method of mathematics morphology. Finally, the descriptor is computed using the scale invariant feature transform (SIFT), with the detected MSCR regions as input. The matching result is obtained according to the nearest neighbor matching based on Euclidean distance of SIFT descriptors. The experimental results prove that this algorithm wins high recognition accuracy and robustness against non-linear image intensity transformation, a substantial range of affine distortion and changes in 3D viewpoint. Also we compare our algorithm to the global appearance based method, and show through experiments in both indoor and outdoor environments that our approach performs better.