Research on SIFT Algorithm Based on Local Color Histogram
Haili Ren · Journal of Guizhou University · 2015
SIFT algorithm is a kind of local feature image matching operator of invariance for rotation,scaling and illumination,which is recognized as one of the best algorithm in the recognition rate. The recognition rate decreases to color targets because SIFT algorithm just uses luminance information and ignores the color information. A novel SIFT descriptor was proposed based on local color histogram( CH-SIFT),in combination with the features of histogram maintaining good rotation,scaling and blur invariant. The locations of feature points of SIFT algorithm generate not only the gradient histogram feature description but also a color histogram feature description. Firstly the match pairs use the gradient histogram feature description to filter,then use the color histogram feature description to filter again,finally determine whether to success under the condition of match. The experimental comparison shows CH-SIFT algorithm has the advantages of high recognition rate and short matching time,which can effectively match the color objects.