[Digikam-devel] Fwd: Re: [CRW-DECODE] The "four colour RGB" option of dcraw

Thorsten Schnebeck thorsten.schnebeck at gmx.net
Tue Apr 18 08:48:34 BST 2006


Hi,

some interesting infos about dcraws -f option

Bye

  Thorsten

----------  Weitergeleitete Nachricht  ----------

Subject: Re: [CRW-DECODE] The "four colour RGB" option of dcraw

Hi Ture,

> Later, I realized that the .CRW files actually contain different white
> balancing values for the two green channels, but dcraw does not use
> this information.

     Starting with v8.15, dcraw will use these separate green
values.  I didn't use them in the past because in one of my
sample images, these multipliers are very wrong.

     "-f" blurs detail and forces dcraw to use VNG instead
of the superior AHD algorithm, so the best option is to white-
balance the greens separately without using "-f".

     Sometimes the ratio of the two greens varies across the
image.  In that case, you have to use "-f".

				Dave Coffin  4/18/2006

On Mon, Apr 17, 2006 at 09:34:17PM +0200, Ture Pålsson wrote:
> Hello List,
>
> Could someone please explain to me the effect of using the "four
> colour RGB" option of dcraw? I understand roughly what it does -- keep
> the two green channels separate through the process -- but not the
> effects of this on the image. Any pointers to explanations that are
> easier to read than the dcraw sources would also be greatly
> appreciated.
>
> The background to my question is the following. I have noticed that,
> without the -f switch, images from my EOS10D tended to have little
> squary patterns in them. This was usually not very pronounced in the
> full-size image, but became obvious when looking at downsampled
> versions, especially when downsampling with a poor filter. Thus, I
> used -f and was happy.
>
> Later, I realized that the .CRW files actually contain different white
> balancing values for the two green channels, but dcraw does not use
> this information. When I modified dcraw to balance the greens
> separately, the "squaries" went away, even when not using the -f
> option.
>
> So, what would be the best way to use? Dcraw-style balancing and -f,
> separate balancing and -f or separate balancing without -f? I have to
> admit I that I do not _see_ any actual difference, but I am curious
> abuot what Theory says. Trying to follow the vng_interpolate source
> code always makes me slightly dizzy, so I tought I would try the easy
> way out and ask...
>
>   -- Ture Pålsson <ture at lysator.liu.se>
>




More information about the Digikam-devel mailing list