Needing insight into the stitching pipeline.
I'm a beginner in OpenCV and I have a few points I dont really understand while reading the source code on image stitching: A big thank you to anyone who could provide some explanation
1) Is remap much faster than warpAffine & warpPerspective and why? I have used warpPerspective on a 3600 times 2000 picture and it is very slow. Same problem with warpAffine, albeit a little faster (it does not perform division I assume). Different story with the cylindrical warper in stitching_detailed: it is not even an algebraic operation but it turns out to be much faster with the same pic. Is it because remap() (which the warper uses) takes 2 pre-computed lookup tables?
2) About 360 panoramas: If I use the cylindrical warper to transform n pics taken with a rotating camera and then use seamFinder, Blender, etc. to compose them into a 360 panorama. Is the final pic a circular band? Or is it still a "filled rectangle" as with any regular pic (I havent tried to compose one yet so cant tell by myself). Lets say the final 360 pano is a filled rectangle. Would't something seem a little off? I mean the left side of pic #1 and the right side of pic #n are supposed to joined together since they represent the angle 0 and 360. But they are not?