Feature Matching of Images

SIFT method for the local description of images is introduced in this paper. After figuring out the various features of the images, this method is used to perform accurate comparison between different views of a scene or an object. The extracted characteristics are invariable to rotation of size, additional noise, change of light, cropped images, etcetera. All the keypoints features of the images are determined from the SIFT Algorithms. For feature matching of the images to check the accuracy, the KNN matching algorithms was used. The approach used in KNN matching is to match the distance ratio from closest point to the next nearest neighbors. A ratio of less than 0.6 is considered as true keypoints and above 0.6 is considered as false keypoints. Thus the main motive to use KNN matching instead of SIFT matching is to increase the accuracy of the keypoints and the matched pairs. The algorithm put forward is to make use of SIFT in real-time application for object detection.