Efficient SIFT descriptor via color quantization

In order to reduce huge computational burden and to improve poor matching performance for existing color SIFT descriptors, we propose a novel color SIFT descriptor via color quantization (CQ). Firstly we quantize the three components of the HSV space and define a one-dimension color vector with a rigorous constraint. Then we introduce the vector to the standard SIFT descriptor and propose the novel color SIFT, termed CQ-SIFT. Experimental results indicate that the proposed color SIFT descriptor is superior to HSV-SIFT and RGB-SIFT.