基于二维最大熵阈值分割的SIFT图像匹配算法

提出了一种基于二维灰度直方图最大熵阈值分割的SIFT图像特征匹配算法。与传统SIFT算法相比,该算法首先综合利用图像像素的灰度信息和邻域空间信息,生成图像二维灰度直方图,并基于此直方图的最大熵对图像进行阈值分割,然后检测分割后图像的DoG尺度空间局部极值,并以此作为特征点进行图像匹配。实验结果表明,基于所提出的匹配算法,可以有效降低图像背景噪声和边缘像素点对目标匹配的干扰,进而提高图像目标的匹配性能。

SIFT Image Matching Algorithm Based on Two-dimensional Maximum Entropy-aided Threshold Segmentation

It is proposed a novel scale invariant feature transform (SIFT) image feature matching algorithm based on the 2-dimensional (2-D) maximum entropy (ME)-aided threshold segmentation. Different from the conventional SIFT algorithm, with the help of the pixel gray level information and neighborhood space information, firstly a 2-D gray histogram is constructed from the raw images and then the image segmentation is processed by the ME of this gray histogram. Secondly, the local extreme value of Difference-of-Gaussian (DoG) scale space of the segmented image is introduced as the feature points for the image matching. Finally, based on the experimental results conducted in the real environments, the image matching algorithm introduced in this paper can be used to effectively reduce the interference of the background noise and edge pixels, and thereby improve the matching performance for image targets.