color image stretching

Re: color image stretching

Martin Leese wrote:
> 
> Is it dull because the intensity is low, or
> because the bands are highly correlated and
> the image consists of subtle shades of grey?
> 
> If the former then you simply need a linear
> contrast stretch, *but* ensure that the
> ratio of red:green:blue remains unchanged.
...

It occurred to me that what I wrote was a
little cryptic, so here is how you do it:

1. Linear contrast stretch the Red band on
    its own, pretending that it is a single
    (graytone) band.  Note the two intensities
    you stretch between (lets call them Rmin
    and Rmax).  Discard the resulting image;
    you only need the two numbers.
2. Linear contrast stretch the Green band on
    its own.  Note Gmin and Gmax.  Discard
    the image.
3. Same for Blue.
4. Start again by applying the same linear
    contrast stretch to all three bands.  The
    two intensity values you stretch between
    are MIN(Rmin, Gmin, Bmin) and
    MAX(Rmax, Gmax, Bmax).
5. Save the contrast stretched image.

I should point out that the normal practice
is to stretch the three bands individually,
and not bother to keep the colours the same
as in the original unstretched image.

-- 
Regards,
Martin Leese
E-mail: please@see.Web.for.e-mail.INVALID
Web: http://members.tripod.com/martin_leese/

Re: color image stretching

On Nov 9, 12:57=A0am, Martin Leese 
wrote:
> Martin Leese wrote:
>
> > Is it dull because the intensity is low, or
> > because the bands are highly correlated and
> > the image consists of subtle shades of grey?
>
> > If the former then you simply need a linear
> > contrast stretch, *but* ensure that the
> > ratio of red:green:blue remains unchanged.
>
> ...
>
> It occurred to me that what I wrote was a
> little cryptic, so here is how you do it:
>
> 1. Linear contrast stretch the Red band on
> =A0 =A0 its own, pretending that it is a single
> =A0 =A0 (graytone) band. =A0Note the two intensities
> =A0 =A0 you stretch between (lets call them Rmin
> =A0 =A0 and Rmax). =A0Discard the resulting image;
> =A0 =A0 you only need the two numbers.
> 2. Linear contrast stretch the Green band on
> =A0 =A0 its own. =A0Note Gmin and Gmax. =A0Discard
> =A0 =A0 the image.
> 3. Same for Blue.
> 4. Start again by applying the same linear
> =A0 =A0 contrast stretch to all three bands. =A0The
> =A0 =A0 two intensity values you stretch between
> =A0 =A0 are MIN(Rmin, Gmin, Bmin) and
> =A0 =A0 MAX(Rmax, Gmax, Bmax).
> 5. Save the contrast stretched image.
>
> I should point out that the normal practice
> is to stretch the three bands individually,
> and not bother to keep the colours the same
> as in the original unstretched image.
>
> --
> Regards,
> Martin Leese
> E-mail: ple...@see.Web.for.e-mail.INVALID
> Web:http://members.tripod.com/martin_leese/

Rmin, Rmax: does it mean Min and Max value in Red band? (or)
Lower and Upper cutoffs for a particular percentage stretch on Red
band?.

>            what is the process of stretching a dull color (NCC,24bit )
> image  from a satellite?, so that it visual appearance will be good,
> but the colors should remain same as original.
>            What i thought is, transform it to IHS color space and
> accommodate the Intentsity component between 0 & 255.
>            When i did this, I got a brighter image than the original,
> but color component also got affected.
>             Can any one suggest me a better method / algorithm.

I don't know what NCC means here but with remote sensing images you
often get quite good visual results just using percent tail trim on each
band separately. Percent tail trim means here a linear stretch between
cut values found from a cumulative histogram by cutting e.g. 1-3 % of
the highest and lowest values. If there are zero filled border areas in
your image, these have to be excluded from the histogram.

You can test this with OpenEV, also with the 16-bit images you mentioned
earlier if they are in tiff format. See Edit->Preferences: Raster:
Autoscaling method and use button "Linear Stretch/Enhancement". You can
get OpenEV in FWTools kit

http://fwtools.maptools.org/

Juho