Temporal-Stereo Depth Estimation

Moving blocks

The demonstrations in these files are of the area/edge based algorithm developed in the at the University of Sheffield as an alternative to the PMF algorithm developed at AIVRU [3]. This algorithm is directly suited to fast hardware implementation and a custom VLSI chip, known as the VCP (Video Convolution Processor) was designed and built for this purpose [4][6]. The algorithm is effectively edge based as it matches data from edge enhanced images but the extraction of edges is postponed until the final stages of the algorithm making the intermediate data structures suitable for maipulation on fine grained parallel hardware, such as [7].
As well as ordinary stereo processing, the stretch correlation algorithm can be used for analysing temporal sequences of stereo images. The files below contain demonstrations of the algorithm together with a sequence of stereo image data and our 3-D results. Implementation of the stretch correlation algorithm aimed to improve on the performance of the standard algorithm by using disparity feedback to guide the feature matching stage from one frame to the next. By seeding the epipolar search stage of the algorithm with the match locations from the previous frame, the algorithm can work with much smaller search ranges, improving robustness by reducing the likelihood of mismatches.
The automated stereo camera calibration algorithms, which enable camera geometery (epi-polars) to be recovered essential if rectification is to be performed, are also available from this link.

Library toolkit
binary 1
Sun Sparc - Solaris 2
binary 2
Intel x86 - Linux 2
[1] P.Courtney, N.A.Thacker and C.Brown. `Hardware Support for fast edge based Stereo.' Proc. European Conference on Computer Vision. Santa Margherita Ligure May, 902-906, 1992.
[2] N.A.Thacker, P.Courtney. `Statistical Analysis of a Stereo Matching Algorithm.' Proc. British Machine Vision Conference. ,Leeds, 316-326, 1992.
[3] R.Lane, N.A.Thacker and N.L.Seed, `Stretch Correlation as a Real-Time Alternative to Feature Based Stereo Matching Algorithms.' Image and Vision Computing, 12,4, pp 203-212, 1994.
[4] R.A.Lane, N.A.Thacker and P.A.Ivey, `A Correlation Chip for Stereo Vision.' Proc. BMVC, York, Sept. 1994.
[5] R.Lane, N.A.Thacker, N.L.Seed and P.A.Ivey, `A Generalised Computer Vision Chip.' accepted for Real Time Imaging 1995.
[6] R.B.Yates, N.A.Thacker, S.J.Evans, S.N.Walker, and P.A.Ivey. `An Array Processor for General Purpose Digital Image Compression.', IEEE Journal of Solid State Circuits, 30, 3, 228-234, 1995.
[7] S.Crossley, A.J.Lacey, N.A.Thacker and N.L.Seed,`Robust Stereo via Temporal Consistency', Proc. BMVC 97, 659-669, Essex, 1997.
[8] A.J.Harris, N.A.Thacker and A.J.Lacey,`Modelling Feature Based Stereo Vision for Range Sensor Simulation', Proc. of the European Simulation Multiconference, 417-421, Manchester, 1998.
[9] S.Crossley, N.A.Thacker and N.L.Seed, `Benchmarking of Bootstrap Temporal Stereo using Statistical and Physical Scene Modelling', Proc. BMVC, 346-355, Southampton, 1998.
[10] A.J.Lacey, N.A.Thacker, S.Crossley and R.B.Yates, `A Multi-Stage Approach to the Dense Estimation of Disparity from Stereo SEM Images', Image and Vision Computing, 16, pp 373-383, 1998.
the end
[Back to software]