A Binary SIFT Matching Method Combined with the Color and Exposure Information

Abstract

The traditional SIFT method is capable of extracting distinctive feature for image matching. However, it is extremely time consuming in the SIFT matching due to the use of Euclidean distance measure. Recently, several binary SIFT (BSIFT) methods have been developed to improve the matching efficiency, whereas merely image brightness information is involved in these algorithms. The matching performance will drop because of the lack of the color information of the image. This paper presents a binary SIFT matching method combined with the color and exposure information. First, three components, including luminance, color offset and exposure offset are combined together to express the image pixel. Then, 128-D SIFT descriptor is converted into 256-bit binarized SIFT descriptor. Finally, the improved Hamming distance is proposed in the matching procedure. Experimental results on UKBench dataset show that the proposed method not only ensures the matching speed, but also improves matching accuracy.