一种多尺度特征点探测算子

A Multi-Scale Feature Point Detector

讨论了传统的图像特征提取算子Frstner算子;提出了一种结合Frstner和多尺度空间的尺度不变特征提取算子,利用图像熵来代替高斯尺度空间的尺度,并进一步通过最大极值熵的方法求图像的局部尺度。实验证明改进后的算子在保持Frstner算子本身优点的同时,又具有尺度不变性。

Forstner operator-a traditional image feature extraction operator was discussed, and an improved method of Forstner operator based on the theory of scale space was establised. The method uses image entropy to replace the scale parameter of Gaussican scale space, and further to calculates the local scale of image by the maximum extremum entropy. This mehtod not only maintain the advantages of Forstner operator but also have the property of scale-invariant.