Match two images using SIFT and RANSAC

Description

SIFT_MATCH demonstrates matching two images based on SIFT features and RANSAC.

SIFT_MATCH by itself runs the algorithm on two standard test images. Use SIFT_MATCH(IM1,IM2) to compute the matches of two custom images IM1 and IM2.

SIFT_MATCH can also run on two pre-computed sets of features.
Use SIFT_MATCH(IM1, IM2, FEAT1, FEAT2), where FEAT1.f and FEAT1.d represent the SIFT frames and descriptors of the first image.

SIFT_MATCH returns MATCHRESULT, where
— MATCHRESULT.RATIO_TEST reports the number of correspondences after the distance ratio test
— MATCHRESULT.RANSAC reports the number of correspondences after the distance ratio test + RANSAC with a homography
— MATCHRESULT.MODEL contains the best homography found
— MATCHRESULTS.MATCHES contains the indices of the matching features in the two images

homography calculation in matlab

Autoencoders of raw images in TensorFlow

https://people.cs.umass.edu/~elm/Teaching/ppt/370/370_10_RANSAC.pptx.pdf