基于概率密度的兴趣点检测算法

针对基于亮度兴趣点检测算法存在对纹理敏感和兴趣点分布不均匀(集中于高对比度区域)问题, 提出了一种新颖的基于概率密度的兴趣点检测算法. 该算法在经典的Harris角点检测算法基础上, 利用亮度的概率密度梯度代替亮度的梯度构建二阶矩矩阵, 进行兴趣点检测. 与基于亮度的Harris检测算法相比, 新算法不仅具有同样的几何不变性, 还有效地抑制了纹理中的``噪声'兴趣点, 并且兴趣点分布的均匀性明显优于基于亮度算法的检测结果.

Density Based Interest Point Detector

To ameliorate the limitation of traditional intensity based interest point detector which is sensitive to texture region a novel density based interest point detector is proposed.The new detector uses the density gradient instead of intensity gradient to build the Harris corner detector.Comparing with the intensity based method,the density based detector can reduce noise corners in the texture region,and the detected corners are distributed uniformly.Thousands of gray level image pairs were used to evaluate the two detectors.The results show that the repeatability of the density based method is nearly equal to that of the intensity based method,but the distribution of density corners is more uniform than that of the intensity corners.