Given two images, e.g. two cats, is there a library that includes a "quick and dirty" way of telling by how much the two images differ regarding translation and rotation? Image registration is a big field and every application I run into seems to be tailored to medical scans and usually has certain domain specific caps on the transformation ranges. The tool I require should take two images as an input and return an angle of rotation and a translation vector, maybe even a confidence metric, it's that simple. (Most algorithms out there are heavy-duty and focus on minute details for alignment, the tool I'm looking for need not be as exact.)
1 Answers
If it needs not be very precise, you can probably tweak the code from PyImageSearch to better suit your application.
If you know that the two images you are going to compare do contain the same object (i.e., if there is no additional object recognition problem that comes before this step), then you can maybe try using the ORB detector to find the good keypoints, and then estimate the homography using ViSP
Related