Newsgroups: sci.image.processing
Path: cantaloupe.srv.cs.cmu.edu!das-news2.harvard.edu!das-news.harvard.edu!news2.near.net!MathWorks.Com!europa.eng.gtefsd.com!howland.reston.ans.net!pipex!lyra.csx.cam.ac.uk!warwick!pcl!palmerj
From: palmerj@westminster.ac.uk (Jake Palmer)
Subject: Re: Contrast enhancement in field crop plots
Message-ID: <CxGF4F.Ewz@westminster.ac.uk>
Organization: University of Westminster
X-Newsreader: TIN [version 1.2 PL2]
References: <HURRELLG.6.000A326A@AgResearch.cri.nz>
Date: Mon, 10 Oct 1994 11:12:14 GMT
Lines: 42

Geoff Hurrell (HURRELLG@AgResearch.cri.nz) wrote:
: We are having difficulty getting sufficient video  colour separation 
: (contrast) between cereal crop plants, weeds and the soil surface.  You would 
: think green plants and brown soil would separate OK; not so.  Soil is 
: remarkably reflective and uneven in colour because of cracks, clods and 
: organic matter on the surface.

: We have tried colour filters, polarizing filters, wetting the soil, artificial 
: light, shade  etc with some success but we need improvement.

: The best so far is:
: -shaded plots or clouded sky, ie. no direct sunlight on the soil surface
: -wet the soil but dry the plant leaves (drops of water are refelective)
: -no filters

: Any suggestions would be appreciated
This is, I think a common problem in imaging of all kinds, video, photographic
and remote sensing, especially where one is attempting to get the computer
to determine where the objects are.  Geoff is right to look at colour 
differences in the scene, however colour imaging is in most cases surprisingly
unproductive and wasteful of computing resources.

Here at the Imaging Technology Research Group we have been perfecting a 
technique for optimising the system to generate maximum contrast in the 
image between chosen objects in the scene.  This can be of benefit
whatever subsequent processing is done by the computer, as the 
maximum information is recorded in the first instance.  The technique
is also applicable, and works admirably, where scene colour differences
are very small (Delta E* < 2).

If anyone has a similar problem of insufficient image contrast
for the image processing task, or wishes to reduce the computational 
complexity and uncertainty in an image segmentation/classification
algorithm, be it process control, computer vision, forensic science,
medical imaging, remote sensing, etc.  then please contact me:
Dr Jake Palmer
Image processing Research Group
University of Westminster
tel +44 (0)171 911 5000 x 2765
fax +44 (0)171 911 5127
snail 18-22 Riding House Street London W1P 7PD, UK
email palmerj@wmin.ac.uk
