Implementation of Stereo Vision Engine

Student thesis: Master thesis (including HD thesis)

  • Tomas Brandt Trillingsgaard
4. term, Signal Processing and Computing, Master (Master Programme)
In this report, a fast high-precision stereo vision engine for implementation on an FPGA is studied. Different aspects of stereo vision and obstacles within this area are explored. Two stereo vision al- gorithms: Efficient Edge Preserving Stereo Matching and Fast Cost-Volume Matching have been studied and implemented in Python. A discussion concerning the al- gorithms computational complexity and stereo matching quality results in the choice of Efficient Edge Preserving Stereo Matching for further implementation. A fi- nal implementation was not achieved but the challenges of implementing an expo- nential function and the challenges with memory usage with large images were studied and solutions were found.
Publication date16 Dec 2016
Number of pages86
External collaboratorHSA Systems
Thomas Heder
Information group
ID: 245525930