To reconstruct point geometry from multiple images, computation of the fundamental matrix is always necessary. With a new optimization criterion, i.e., re-projective 3D metric geometric distance rather than projective space under RANSAC (Random Sample And Consensus) framework, our method can reveal quality visually through reconstruction. The projection error points to corresponding image pixel...

This paper derives a direct constraint satissed by points on the axis of articulation of a compound object. A minimum of three views of the articulated object in motion are required to recover the projective position of the articulation axis. No camera calibration is required. The method given here is a signiicant improvement on a previous method, doing away entirely with the need to use the fu...

Transparent refractive objects are one of the main problems in geometric vision that have been largely unexplored. The imaging and multi-view geometry of scenes with transparent or translucent objects with refractive properties is relatively less well understood than for opaque objects. The main objective of our work is to analyze the underlying multi-view relationships between cameras, when th...