HSI and HSV
du97
the difference between HSV and HSI colour spaces? (Subject: HSI and
HSV, 1995).
Some replys stated that there was little difference and other messages
directed readers to usefull code for converting RGB to HSV:
http://www.cs.rit.edu/~ncs/color/t_convert.html
http://homepages.borland.com/efg2lab/Graphics/Colors/HSV.htm
I have found that there is a diffence between the V (Value) and I
(Intensity) calculations and that this had an adverse effect on my
work.
I found that V = max[R,G,B] and that I = (R+G+B)/3.
I am working on a method that I originally applied to the I band of
the HSI space. Result were good. In converting my method to a Windows
based system I started to use a commercially avalible method that used
HSV. Assuming that HSV was the same as HSI I was supprised when my
results were not as good. The problem was that in our primary images
the red channel dominates (has the highest values) but shows little
contrast for the important features. Processing the V band proved not
as successful as working with the I band.
So there is a difference between HSV and HSI and you have to choose
the most appropriate for your circumstances.
P.S. In viewing the images using Photoshop the HSB band that is used
appeared similar if not the same as the HSV representation.
Will Dwinnell
du...@email.com (du97) wrote:
"I have found that there is a diffence between the V (Value) and I
(Intensity) calculations and that this had an adverse effect on my
work.
I found that V = max[R,G,B] and that I = (R+G+B)/3."
This raises an interesting question: are transformations to either of
these colorspaces truly reversible? The maximum function surely
destroys information and, I am guessing, makes perfect recovery of the
original color data impossible (?). What about HSI? Are any of the
usual "separate luminance and chrominance" transforms (YCbCr, etc.)
perfectly reversible?
-Will Dwinnell
http://will.dwinnell.com
du97
> This raises an interesting question: are transformations to either of
> these colorspaces truly reversible? The maximum function surely
> destroys information and, I am guessing, makes perfect recovery of the
> original color data impossible (?). What about HSI? Are any of the
> usual "separate luminance and chrominance" transforms (YCbCr, etc.)
> perfectly reversible?
>
Regarding the reversiblity of the HSV system although the max function
destroys information it may be possible to recover which RGB value was
maximun using the hue angle?, i.e. which colour axis is closest.
Gernot Hoffmann
HSV=HSB is reversible. HLS is reversible. I don愒 know HSI.
Code for HSV and HLS is in Foley,van Dam, Computer Graphics, or by
a Google search like "Paul Bourke Color Conversions".
I would recommend HLS (double cone), it愀 more intuitive.
Code for standard HLS and a modification of HLS, which is more appropriate
for image processing, can be found here:
http://www.fho-emden.de/~hoffmann/hlscone03052001.pdf (500kBytes)
I am using either 1/3 weights or common RGB weights, maybe NTSC, as usual.
The original Foley HLS system has severe discontinuities near to the
poles.
HSB=HSV is not intuitive (IMO) because Brightness is Max(R,G,B),of course
normalized. 255 0 0 is as "bright" as 255 255 255 .
Interpolation is hardly possible.
Best regards --Gernot Hoffmann