Rectification on Uncalibrated Epipolar Stereo Images and Dense Disparity Map

K. Takaya (Canada)


Fundamental Matrix, Epipolar Geometry, Ransac, Rectifi cation, Disparity Map


This paper presents a result of the studies to generate the dense disparity map based on the epipolar geometry of stereo vision. The distance (missing depth in 2D images) to a point on the objects recorded in a pair of stereo im ages can be estimated from the disparity due to the par allax between two cameras. The well established theories of the epipolar geometry combined with the robust method of RANSAC to calculate the fundamental matrix was im plemented. Stereo images are rectified to make epipolar lines parallel along the horizontal axis (raster lines), and to enable 1D search for matching the points of correspon dence. This paper proposes a scanning type pattern match ing method that can be used to generate a dense disparity map which displays a 2D distribution of pixel-by-pixel dis parities. Preliminary results are presented.

Important Links:

Go Back