Abstract:

In order to avoid the disadvantages of huge computational complexity and poor matching performance in existing  scale invariant feature transform（SIFT) descriptors for color images,  a novel SIFT descriptor for color images based on color quantization matrix was given. The quantization matrix was obtained by triplecolor components. And it was applied to generate a 128dimensional SIFT descriptor to match color objects. The experimental results showed that, compared with existing SIFT descriptors for color images, this proposed method has the advantages of higher correct matching rate, less matching time and more accurate matching points. This method can effectively match the color objects.

Key words: color quantization matrix, scale invariant feature transform (SIFT), hue saturation value(HSV) color space, feature matching