McGill Calibrated Colour Image Database
How to Colour Calibrate the Images
The images can be calibrated in two ways, either (A) to gamma-correct the images, or (B) to gamma-correct the images and convert them from RGB to LMS space. In LMS space the three planes are the modeled responses of the L, M and S cones of a human observer. The RGB and LMS values of the calibrated images are in double precision floating point format, and thus need to be rescaled for displaying on a monitor. The calibrated RGB and LMS values have also been divided by the camera exposure time, which is automatically adjusted each time a picture is taken to ensure that the image falls within the camera's dynamic range. This means that the calibrated RGB and LMS values are linearly related to the absolute luminances of the input image. The exposure time is given for each picture when the calibration routines are run.
The calibration functions are written in MATLAB, and require the images in TIF format. Click on one of the following options to download the required calibration routine. After downloading, uncompress the downloaded file and copy it into your matlab working directory.
Next, in MATLAB, calibrate any image from the database as follows:
>> [LRGB] = rgb2linear('image');
or
>> [LMS] = rgb2lms('image');
where 'image' gives the path and name of the image to be calibrated.
Note that the above routines correct the gammas of the cameras, and are not applicable to any device used for displaying the images.
FURTHER INFORMATION
Full details of how the cameras were calibrated and how the Matlab conversion routines were coded can be found in BACKGROUND INFORMATION.