Hybrid Method of 3-D Image Reconstruction from Stereo Pictures

A hybrid method of 3-D image reconstruction from stereo pictures is introduced. The proposed method utilizes the modified algorithm SAD where Depth map is calculated from 3-D noisy Cost function, which is de-noised by a new de-noising technique with block-matching and 3-D filtering. A technique of Epipolar geometry is then applied to stereo pictures. False corner matches in stereo pictures are discarded by comparing their position with the Depth map. The resulting correct corner matches are used to calculate a final Fundamental matrix F by a linear method and its Error matrix is utilized to correct great errors in the final Depth map. The method successfully improves results of 3-D image reconstruction in comparison with SAD or Correlation algorithms.