基于Hessian矩阵和Gabor函数的局部兴趣点检测

A NOVEL HESSIAN-GABOR BASED LOCAL INTEREST POINT DETECTION

局部特征方法是基于内容的图像与视频检索的重要方法。提出一种新的基于Hessian矩阵和Gabor函数的尺度不变局部特征点检测方法(Hessian-Gabor Detector)。该方法首先利用基于Hessian矩阵的检测子定位图像在空间域上的候选特征点位置,然后用基于Gabor函数的算子检测候选兴趣点在尺度空间的特征尺度,从而获得具有尺度不变特性的局部特征点。实验证明,与DOG、Harris-Laplace等方法相比,计算简单。应用于图像匹配中,能够显著地提高匹配效率。

Local feature is an important method in content-based image and video retrieval.This paper proposes a new scale invariant local feature points detection method based on Gabor function and Hessian matrix（Hessian-Gabor Detection）.This method obtains the position of candidate feature points of the image in spatial domain with the detector based on Hessian matrix at first,and then detects the characteristic scale of candidate interest points in scale space with detector based on Gabor function,therefore the local feature points with scale invariant are attained.Experiments demonstrate that the approach proposed in this paper has simpler computation comparing with the DOG and Harris-Laplace.It can be used in image matching with conspicuously improved matching efficacy.