THIS PAGE IS UNDER CONSTRUCTION
This page provides additional experimental results concerned with the mapping of the Fast Bilateral Stereo (FBS)
approach on GPUs described in this paper 
: S. Mattoccia, M. Viti, F. Ries, "Near real-time Fast Bilateral Stereo on the GPU", Best Paper Award at 7th IEEE Workshop on Embedded Computer Vision (ECVW20011), CVPR Workshop, June 20, 2011, Colorado Springs (CO), USA [this is an updated version of the paper: see the NOTE for details]
Compared to the original CPU implementation of FBS, our
mapping on a medium-class NVIDIA 460 GTX with CUDA allows us to
obtain, with equivalent results, speed-ups > 70X
and on a Tesla C2070 speed-ups > 100X
With the NVIDIA 460 GTX and blocks of size 3x3 the execution time is 65 ms
with the NVIDIA Tesla) on Tsukuba and 302 ms
ms with the NVIDIA Tesla) on Teddy/Cones.
With the NVIDIA 460 GTX and blocks of size 5x5 the execution time is 40 ms
with the NVIDIA Tesla) on Tsukuba and 178 m
ms with the NVIDIA Tesla) on
is an algorithm that combines the effectiveness of
state-of-the-art cost aggregations strategies, that adapt their weights
to image content, with the efficiency of fast incremental
calculation schemes (i.e. integral images or box filtering) typically
conventional stereo matching algorithms.
FBS computes the weights on a block basis and the matching cost on a
this strategy enables, in both cases, to deploy efficient box-filtering
integral images) incremental calculation schemes. Additional
information, experimental results and software concerned with the FBS
approach can be found here
. S. Mattoccia, S. Giardino, A. Gambini, "Accurate and efficient cost aggregation strategy for stereo correspondence based on approximated joint bilateral filtering", Asian Conference on Computer Vision (ACCV2009)
, September 23- 27, 2009, Xi'an, China
is an updated version the paper. The originally published paper
unintentionally failed to describe properly the following paper:
C.Richardt, D. Orr, I. Davies, A. Criminisi, N. A. Dodgson, "Real-time Spatiotemporal Stereo Matching Using the Dual-Cross-Bilateral Grid", ECCV 2010
fact, this algorithm computes weights symmetrically and its loss
in accuracy, compared to the adaptive weights method, comes primarly
from the use of greyscale images.
We thank NVIDIA Corporation for their interest in our research and the donation of a Tesla C2070 GPU.