I need a program that takes two incoming images (original A and warped B), finds high confidence correlation points and warps image B to line up to image A. The warp will in image B will non-linear and variable across the image. Piecewise homography (i.e. OpenCV) or blended piecewise polynomials are the desired solution.
Code to be in C/C++ for Windows using OpenGL (or the "Processing 3" language). Input and output files are uncompressed PNG. Minimal UI: load input images, see image A, image B, a correlation point image, and image B overlayed on image A with adjustable transparency to check results.
If successful, project extension for additional features to be added.