simonfuhrmann
commented
7 years ago The depth maps that scene2pset outputs are already aligned. It uses the camera parameters from SfM to do that. These parameters are usually VERY precise, i.e., in the order of sub-pixel reprojection accuracy. So there is really no point in doing ICP, and I bet ICP is going to make the alignment MUCH worse.
Hello Sir
I was reading kinect fusion paper and they mentioned ICP algorithm and than I remembered that this algorithm is good for association between point clods. in MVE/scene2pset you just project depth maps all together in 3D to have final dense cloud. what if we align the projected depth map using ICP. would that effect the result of fssrecon.