home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!gatech!usenet.ins.cwru.edu!agate!dog.ee.lbl.gov!overload.lbl.gov!maui.lbl.gov!wes
- From: wes@maui.lbl.gov (Wes Bethel)
- Newsgroups: comp.graphics.visualization
- Subject: (Re: Histogram Equalization) image processing of multi-band images
- Followup-To: comp.graphics.visualization
- Date: 31 Aug 1992 16:01:41 GMT
- Organization: Lawrence Berkeley Laboratory
- Lines: 49
- Message-ID: <17tfp5INN1fl@overload.lbl.gov>
- NNTP-Posting-Host: maui.lbl.gov
-
-
- In article <1992Aug26.205039.7555@rock.concert.net> krueger@rock.concert.net (Richard C Krueger -- Fauve Software) writes:
- >I am trying to equalize a 24 bit color image. I have no
- >problem with 8 bit grayscale images. I have tried equalizing
- >each channel (R, G, and B) independently, but the results are
- >not what you would expect. Has anyone out there solved this
- >problem. Help or pointers to the right articles would be
- >greatly appreciated.
- >
- >Richard Krueger.
-
- We had a related problem here - we have a Kodak XL7700 printer which
- tended to print images much "too dark."
-
- I began to investigate using gamma correction as a built-in part
- of the print driver. The thing I discovered, as I'm sure anyone
- who has tried "image processing" of 24 bit images has discovered,
- that by applying an intensity-modifying operator of some type
- (histogram eq, gamma correction, etc.) independantly to each color
- channel that one can get a really obnoxious hue shift.
-
- To make a long gripe short, this was not OK.
-
- One thing I tried was to convert from RGB space to HSV space,
- and applying the operator to the V, or "brightness" component.
- This worked OK in that the hues were preserved (e.g., brown didn't
- turn purple), but was not completely satisfactory for my application
- which was to provide an image from the printer that looked
- "the same as" the one the scientist was looking at on their
- monitor.
-
- This raises the ugly head of device calibration which I wish
- to avoid for the moment.
-
- It seems that this process (convert to HSV, histogram eq the V,
- then convert back to RGB) may work OK for histogram eq'ing
- 24 bit RGB images.
-
- Try it and let us know how it turns out, ok?
-
- As for us, it turns out that applying the gamma correction to each channel
- independently IN GENERAL gave more satisfactory results than when tying
- to preserve the hue and saturation and tweaking brightness only.
-
- "Satisfactory" is a subject metric, of course, so your mileage will vary.
-
- wes
- ewbethel@lbl.gov
-
-