On this page a selection of the images used for machine vision research
at the University of Manchester and the University of Sheffield are made available for you to download.
The stereo image pairs are made available in both aiff (TINA
image file format) and pgm (portable greylevel map, readable by
many packages including XV). All stereo pairs have
the filename structure : basename.l.type and basename.r.type.
Where l refers to the left image, r refers to the right image
and type is either aiff or pgm. The stereo sequences
have an added extension denoting the frame number thus: basename.xxx.l.type.
Where xxx is the 3 digit frame number starting at 000. The remaining images are single (mono) images and have the extensions aiff or pgm.
None of the aiff images were captured using parallel stereo camera systems
so for the aiff images epipolar / raster alignment is not guaranteed. Using
the TINA stereo calibration software, the geometries for the camera systems
involved were derived and from that information, camera (.cam) files
have been generated which are supplied with each pair of aiff images.
In the TINA environment, the camera files allow the stereo vision tools
to work with the aiff images directly, and return results in mm. If your
stereo algorithms cannot handle arbitrary camera geometries, a corresponding
set of pgm images are supplied which have been rectified using TINA so
the epiolar lines for stereo pairs are aligned with the image rasters.
The images in the library are all compressed using gzip and require gunzip to remove them.
Where there are multiple images these have been grouped and compressed
using tar and gzip. To recover these images it is necessary
to use gunzip followed by tar -xvf <filename>.