In this project, we calculate the disparity map of two rectified images which imitate two eyes. In this project, SAD and DP algorithms (simplified) are implemented. PyCUDA codes are typically included.
- "cudadpAdvanced.py" is the pycuda code for Dynamic Programming.
- "sad.py" is the pycuda code for local algorithms (SAD&SSD).
- "stereo-matching.py" is local algorithms which could be implemented on CPU.
- "dpmatching.py" is dp algorithm which could be implemented on CPU.
- "imL.png" and "imR.png" are input left images and right images.
Execution of dp algorithm and sad are around 0.2-0.4s on GPU and 10^-3 matching. We also enclosed disparity maps we derived.
This is a code by Qingwei Wu and Reixuan Zhang
Please feel free to contact me: qw2208@columbia.edu