Image Correspondence Using Affine SIFT Flow

摘要

Image correspondence is one of the critical tasks across various applications of image processing and computer vision. The simple correspondence is being studied from many years for the purpose of image stitching and stereo correspondence. The images assumed for the simple correspondence have same pixel value even after applying the geometric transformations. In this work, we try to correspond images sharing similar content but vary due to change in acquisition like view angle, scale, and illumination. The use of features for flow computation was proposed in SIFT flow, which was used for correspondence across fields. In this method, dense SIFT descriptors are extracted and flow is estimated for matching the SIFT descriptors between two images. In this work, we applied SIFT flow algorithm on the affine transformations of images to be aligned.