关键词: 图像匹配, SIFT, 对应尺度LTP, 相对灰度直方图, 街区距离

Abstract:

SIFT is one of the most robust and widely used image matching algorithms based on local features. However, the computational and matching complexity of the descriptor with 128 dimensions is too high. In order to speed up image matching, an image matching algorithm is proposed. Firstly, measures such as constructing scale spaces, detecting extrema and confirming keypoints are carried out to create the key feature points. Secondly, the key feature points are described by the rotation-invariant LTP which is formed by the patches around the key-points, and by the relative grayscale histogram. Lastly, rotation invariant descriptors are formed, and the city-block distance is adopted for image matching. Experimental results prove that the robust method improves the matching speed and matching precision even under the condition of illumination transformation, scale transformation, and rotation transformation.

Key words: image matching;SIFT;corresponding scale LTP;relative grayscale histogram;cityblock distance