Intrinsic Image Benchmark - Synthetic St. Basil - v2.0, July 2013

The original version of this benchmark (v1.0) contains the following files:
- 'input_singleImageMethods' folder: 
	o rendered images for 3 viewpoints in linear, high-dynamic range, .EXR format
	o rendered images converted to .PNG (see conversion formula below)
- 'groundtruth' folder:
	o ground truth illumination images, computed with path tracing (in .EXR and .PNG formats)
	o ground truth reflectance images, obtained by dividing input and illumination images (in .EXR and .PNG formats)
	o binary mask where sky pixels have been set to 0 (in .PNG format)

All linear .EXR images were converted to .PNG with the following formula: imgPng = (alpha * imgExr) ^ (1/gamma)
Parameter alpha was chosen so that at most 2% of pixels within mask are saturated:
- alpha = 0.7 	for input images,
- alpha = 1 	for reflectance images,
- alpha = 0.29 	for illumination images,
- gamma = 2.2 	for all images.	
	
The extended benchmark (v2.0) contains in addition the following files:
- 'input_fixedViewpointMethods' folder:
	o rendered images from a fixed viewpoint facing the St. Basil cathedral. 30 realistic lighting conditions are used. Images are in linear, high-dynamic range, .EXR format
- 'input_multipleViewpointMethods' folder:
	o rendered images from 30 viewpoints, with 30 realistic lighting conditions. Images are in linear, high-dynamic range, .EXR format
	o camera calibration file, in Bundler format (see documentation at http://www.cs.cornell.edu/~snavely/bundler/bundler-v0.4-manual.html)
Results can be compared to the ground truth reflectance for viewpoint 101, which is named 'imageA' in the 'groundtruth' folder.
	