I am quite new in Vision and Image Processing. I am trying to find the best toll to estimate the 3D motion vectors of the interest points of an object, having the 2D motion flow. In another word, I want to recognize the 3D motion of an object, having a sequence of 2D images taken from a standing still camera. What would be the best tool to convert these 2D vectors to 3D ones, without knowing the object geometry. (We might only know that the object is rigid, and hence the 3D distance between the interest points will not change.)
Many thanks in advance