An Improved SIFT Descriptor and Its Performance Analysis

In order to solve the problems that conventional SIFT descriptors result in a lot mismatches when an image has many similar regions,an improved SIFT Descriptor and its generation method are proposed.Firstly,the steps of constructing scale space,detecting extrema and confirming keypoints are employed to create the key feature points.Secondly,each keypoint is described by the rotation-invariant Texture patterns,which are computed from the image patch centered at the keypoint,and be combined to SIFT feature to construct the union feature.Finally,a rotation invariant textured SIFT descriptor is formed.Two different matching strategies are adopted as the method applied to image matching.The experiment show that the method makes the image content information more comprehensive,it enhances the accuracy of matching and improves matching results greatly,achieve the improvement of SIFT algorithm.