[Image-SIG] Re: Returning Colormap values for an image?

[Alfred Milgrom]

Hi:

Thanks for your reply.

The use of the additional parameters in 
im = im.convert("P", dither=Image.NONE, palette=Image.ADAPTIVE)
has made all the difference!

My original image had only 3 colours, but using the plain vanilla
im.convert("P") construct resulted in an image with 6 colours!

However, with the additional parameters the conversion is perfect.
Thanks again,

Alfred Milgrom

On Mon, 23 Aug 2004 13:46:44 +0200, Fredrik Lundh
<fredrik at pythonware.com> wrote:
> Alfred Milgrom wrote:
> 
> > The only problem (if you can call it that?) is that the colours
> > actually used in the original image are not necessarily the first n
> > values in the lut image, and so it is necessary to not only work out
> > how many colours are used initially but also which indices are used.
> >
> > Is there an easier way to get a list of the indices used than something like:
> >
> > hist = im.histogram()
> > coloursUsed = [i for i in range(len(hist)) if hist[i]]
> 
> not really: the histogram approach is the only efficient way to find out
> what colors you have in an L or P-mode image.
> 
> > Also, if the image I have is an RGB image (such as the result of an
> > ImageGrab), what is the best way to get the colours used? Is
> > converting to 'P' mode and then using the resize/convert method the
> > easiest/fastest?
> 
> probably, as long as you're aware that you won't get exact results if
> the image contains more than 256 colors.
> 
> for best results, use:
> 
>     im = im.convert("P", dither=Image.NONE, palette=Image.ADAPTIVE)
> 
> (no dithering, adaptive palette.  if you need more control over the
> process, you can use the undocumented quantize method, but you'll
> have to read the source code to figure out how that works).