For all the datasets of different cameras, FIVE types of files are provided:

RAW images:

	The original camera dependent RAW images. You can reprocess these RAW images
	if you need them for other purposes.


JPEG images: 

	The original camera default output JPG images, which are given to show the
	content of each image. To be noted, the JPG image size is slitly smaller
	than the according RAW image size due to the cropping process inside the 
	camera.


PNG images:	

	Our proccessed images files using the DCRAW software. Arguments '-D -4 -T'
	was used to just extract the original linear image data. No further 
	demosaicing was done, but only downsmapling according to bayer pattern to 
	create a color image. Color patterns of all 8 cameras, except the Fujifilm 
	X-M1, are all variations to the standard RGGB Bayer patter, so the original 
	DCRAW coverted images were downsampled by factor 2. The pattern used by the 
	Fujifilm X-M1 is a noval 6*6 pattern, which is termed X-Trans by Fujifilm,
	but in every 3*3 block, there exists RGB, so the images from Fujifilm X-M1
	camera were downsampleed by factor 3.


Color-Checker information files: 

	The *_CHECHER.zip file contains the color checker information including the 
	mask positions and colors of all the patches. There are two files for each 
	image: 

	*_mask.txt	
	-- 	Includes the ROIs of the color-checker as well as all 24 patches.
		The first line is the ROI of the color-checker while the next 48 
		lines are respectively x-coordinates and y-coordinates of four
		warping rectangle  corners of 24 patches. 
		For example, the first 3 lines maybe:
			92.318,1304,591.65,425.28
			35.232,35.452,91.167,90.676
			317.83,369.93,367.37,315.28
		Use the following matlab code to extract the color patch:
			roi = [92.318,1304,591.65,425.28];
			xm  = [35.232,35.452,91.167,90.676];
			ym  = [317.83,369.93,367.37,315.28];
			Icc = imcrop(image,roi);
			pch = poly2mask([xm xm(1)], [ym ym(1)], size(Icc,1), size(Icc,2));

	*_color.txt	
	-- 	Mean colors of all the 24 patches inside each ROI defined 
		in its *_mask.txt file, starting from bottom to top and left to
		right, so that the achromatic patches are the first six.


Groundtruth .mat file:

	A matlab matrix file that contains necessary information for processing all 
	the images, including groudtruth illuminants extracted form color-checkers. 
	The structures of all the groundtruth .mat files are the same:

	all_image_names			
	-- 	all the image names	

	groundtruth_illuminants	
	-- 	the according groudtruth illuminants of all the images

	CC_coords				
	-- 	the coordinates of the color-check in the image
		the upper left corner is [CC_coords(1), CC_coords(3)]
		the lower ight corner is [CC_coords(2), CC_coords(4)]

	darkness_level          
	-- 	use with saturation_level to normalize the image (must be subtracted 
		from the original images before any usage)

	saturation_level		
	-- 	use with darkness_level to normalize the image and detect regions 
		possibly saturated (to detect the possibly saturated pixels, a 
		threshold less than this saturation_level is recommended, like 98% of 
		the saturation_level, as the estimation of this saturation_level may 
		not be accurate)