Implementation of stereo correspondence algorithms for multi-baseline vision system
Department of Biomedical Engineering
Master of Science
Foulds, Richard A.
Simone, Lisa K.
The purpose of this study is to determine the impact of varying camera baselines on correspondence and depth estimation. As the baseline increases the resolution of depth increases but finding correspondences is very tedious. On the other hand when cameras are placed close to each other and thereby reducing the baseline distance points in the images are easily matched however depth resolution deteriorates. This problem is solved by using a multi-camera system is used in which the cameras are spaced close to each other along a straight line. Calculating the correspondence along successive camera pairs is relatively easy. These respondences are later propagated to the pairs with larger baselines.
Another important aspect of this study was to implement different area based stereo correspondence algorithms. Advantages and disadvantages are noted in each case. It is noticed that in some cases simple SSD and SAD worked very well but failed in other cases where Second Order Correlations worked well.
njit-etd2007-012 (77 pages ~ 10,457 KB pdf)
Please complete this Feedback Form to inform us about your experience using this website. It will assist us in better serving your information needs in the future. Thank You!
Created September 9, 2008