Detection of rotation and scale invariant regions using contour

In order to obtain robust feature regions,this paper presented a method using contour to detect image feature regions which were invariant to rotation and scale transformations.Firstly,it applied multi-scale production LoG to extract robust corners on the contour as feature points.Secondly,the method used angle bisector which was invariant to rotation and scale transformations to extract feature orientation and got feature radius by feature orientation.Finally,corners,feature orientation and feature radius were used to construct invariant regions.The test images used in image matching experiments include five kinds of transformation,such as rotation,scale,affine,brightness and compression.The excellent matching results illustrate that the proposed method has the properties such as easy realization,fast,better robustness and wide range of applications.