Path: utzoo!utgpu!jarvis.csri.toronto.edu!mailrus!ncar!tank!james@rover.bsd.uchicago.edu
From: james@rover.bsd.uchicago.edu
Newsgroups: comp.graphics
Subject: Expanding your camera's dynamic range
Message-ID: <3113@tank.uchicago.edu>
Date: 8 May 89 14:55:00 GMT
Sender: news@tank.uchicago.edu
Organization: University of Chicago - Dept. Rad. Onc. and Med. Physics
Lines: 30


	Well, barring the hopefully eventual creation of sci.med.physics (plug!)
I thought this would be the next best forum to open comments on an idea that
someone else may have already tried.

	I have a problem, in that I am digitizing x-ray films with a laser
scanner, an optical drum scanner, and a ccd camera.  The first two modalities
obviously produce better images (re. dynamic range, resolution, and SNR), but
the camera is the most economically feasible method.  Many problems of local
contrast enhancement and noise can be addressed by image processing methods,
but the overall contrast range is limited by my camera and digitizer (8 bits).

	I am now trying a method for expanding the dynamic range.  I capture
two images, one at a low iris setting and one at a high setting, By taking 
histograms of these two images, I find the levels at which useful information
gets "washed out" by either flooding at high intensities or quantum effects
at low intensities.  I take ranges of useful grey values in the two images, 
expand them to the same scale, and add the two images.  I then do linear
histogram equalization to normalize the contrast distribution (which is
no longer linear).  My results are ok, in that I can now see contrast in both
the very bright and very dark regions of the film, but my normalization
inherently enhances noise.  I have tried clipping the histogram, but there
still must be a better way.  My next attempt will be with adaptive histogram
equalization, but I am open to suggestions!



James Balter
James@rover.uchicago.edu
"If the hat fits, slice it!"