Path: utzoo!utgpu!news-server.csri.toronto.edu!rutgers!bellcore!uunet!shelby!leland.Stanford.EDU!hanauma!rick
From: rick@hanauma.stanford.edu (Richard Ottolini)
Newsgroups: comp.graphics
Subject: Re: Medical CT Scan info
Keywords: medical, CT, imaging
Message-ID: <1991Feb28.162456.18328@leland.Stanford.EDU>
Date: 28 Feb 91 16:24:56 GMT
References: <12803@helios.TAMU.EDU>
Sender: news@leland.Stanford.EDU (Mr News)
Distribution: usa
Organization: Stanford University, Department of Geophysics
Lines: 21

In article <12803@helios.TAMU.EDU> jim@sniff.tamu.edu writes:
>
>Hello,
>
>I hope that this is the right group to post to.
>
>using Sunvision on a Sparcstation 1+. Our CT machine produces
>256 x 256, 16 bit images. Some of the problems that we have had is
>in scaling the data to 8 bits for Sunvsision. We have had some luck
>in displaying images, but there seems to be some degradation between
>what the CT machine displays and what we are able to display.  Our

Use a non-linear conversion from orignal data to byte integers.
One way is to partially sort the data into a histogram and scale the existing
range between 0 and 255.  Another enhancement is called the gamma contrast
enhancement.  That is to raise each sample to a fractional power.  This can be
done by table-lookup at both the conversion time and display time.  We use
an ad-hoc partial histogram to obtain a reasonable gamma.  We compute an
upper bound and half intensity value of the data distribution and fit a gamma
power to that.  For seismic data which is distributed towards zero, we use the
99th and 85th percentile for these two values.