Using the GPU for Fast Symmetry-Based Dense Stereo Matching in High Resolution Images
Institute of Systems and Robotics, Dept. of Electr. & Computer Eng., University of Coimbra, Portugal
IEEE International Conference on Acoustics, Speech, and Signal Processing (ICASSP’2014), 2014
@article{mota2014using,
title={Using the GPU for Fast Symmetry-Based Dense Stereo Matching in High Resolution Images},
author={Mota, Vasco and Falcao, Gabriel and Antunes, Michel and Barreto, Joao and Nunes, Urbano},
year={2014}
}
SymStereo is a new algorithm used for stereo estimation. Instead of measuring photo-similarity, it proposes novel cost functions that measure symmetry for evaluating the likelihood of two pixels being a match. In this work we propose a parallel approach of the LogN matching cost variant of SymStereo capable of processing pairs of images in real-time for depth estimation. The power of the graphics processing units utilized allows exploring more efficiently the bank of log-Gabor wavelets developed to analyze symmetry, in the spectral domain. We analyze tradeoffs and propose different parameterizations of the signal processing algorithm to accommodate image size, dimension of the filter bank, number of wavelets and also the number of disparities that controls the space density of the estimation, and still process up to 53 frames per second (fps) for images with size 288 x 384 and up to 3 fps for 768 x 1024 images.
July 12, 2014 by hgpu
