基于差分形态分解的多尺度Harris角点检测器

Multi-scale Harris corner detector based on differential morphological decomposition

为克服Harris-Laplace角点检测器存在的检测精度不高,易受噪声干扰,积分尺度选择困难等问题,提出了一种基于差分形态分解的多尺度Harris角点检测器。其利用差分形态分解得到的图像结合相关尺度信息形态学重建原始图像,以准确定位角点位置。实验结果表明,相比多尺度Harris-Laplace角点检测器,其不仅定位准确,大尺度下图像边缘保持完好,而且误检率低,尤其在角落区域和重叠区域检测优势明显。

In order to overcome the deficiencies of the Harris-Laplace corner detector, such as poor detection accu？racy, noise interference, and difficult selection of the integral scale, this paper proposed a new multi-scale Harris corner detector based on differential morphological decomposition （ Harris-DMD） . This detector combines the image got by the DMD with the related scales information to morphologically reconstruct the original image, so as to accu？rately locate the position of the corner points. The experiments show that, compared to multi？scale Harris？Laplace corner detector, the Harris-DMD improves the detection accuracy, preserves the image edge well, and has a low rate of the false detection. Furthermore, it has a great advantage on corner regions and overlap regions.