One of application of homography is in image processing. For example in character recognition. We have many template images. We need a scoring function, which calculate similarity between input image and template images, Each image in the templates that has lower average inlier score is selected as recognized character.
A match and score computation is done between this character-to-be-identified and each template image. The frames of the matches are then ran through RANSAC to calculate a best-fit homography matrix and identify inlier features between the character and template. RANSAC sets, k, the number of matching pixels needed to compute the homography and samples for the best homography S times.
The average score of these inlying features is taken, and the template which achieves the lowest average inlier score is selected as the character’s identity and match.