I have a question about Dense reconstruction using Thin-Plate Splines interpolation. I have two images, then I detect some features using SURF (sparse reconstruction). These points were used to initialize the interpolation which seems good. So, I have two grids of corresponding points. The problem consists in fact that the triangulation of these grid points gives a very bad result in termes of dense reconstruction... In 3D, I obtain the same 2D grid but in 3D. Unfortunately I can't put images directly in my message, but you can find them here :
Results of sparse reconstruction
Incorrect dense reconstruction
My question is : what am I doing wrong? Should I assign features to new interpolated points and then match them again?
Thank you in advance !
