Ground truth stereo videos
==========================

This groud truth stereo video was created for evaluation purposes for the paper

  Real-time Spatiotemporal Stereo Matching Using the Dual-Cross-Bilateral Grid
  Christian Richardt, Douglas Orr, Ian Davies, Antonio Criminisi and Neil A. Dodgson
  European Conference on Computer Vision (ECCV) 2010

  Project website: <http://www.cl.cam.ac.uk/research/rainbow/projects/dcbgrid/>

If you use our datasets, please cite our paper. Thanks.


Dataset 'book'
--------------
length      : 41 frames (0001 to 0041)
resolution  : 400 x 300
disparities : 0 to 63.75 in steps of 0.25

To obtain disparity (in pixels) from pixel values, devide the pixel value by 4.
This maps pixel values 0 to 255 linearly to disparities 0 to 63.75, in steps of 0.25.

The book object was taken from the Official Blender Model Repository at
<http://e2-productions.com/repository/>.


Naming scheme
-------------
The naming scheme for images in sub-directories is as follows:

  L0001.png  - Left input frame.
  R0001.png  - Right input frame.
  TL0001.png - Left ground truth disparity map.
  TR0001.png - Right ground truth disparity map.