[ptx] Vignetting correction in nona

[JD Smith]

On Mon, 2006-01-02 at 20:15 +0100, Pablo d'Angelo wrote:

> > I'd also suggest normalizing the flat image
> > to the median, rather than the max, to ensure no overall shift in
> > brightness.
> 
> Hmm, but the image should have been brighter, and its brighness at the
> corners was reduced by the vignetting... The question is wether to risk
> clipping pixels or not (in the 8 bit case..).

>From my point of view, the vignetted image was the one metered for.  If
the vignetting is severe, the camera (or photographer) might choose to
up the exposure to compensate.  For this reason, attempting to keep the
median brightness pixel the same brightness (which means some pixels
will get fainter) could be better than boosting everything up even
closer to the digital saturation limit.  For typical vignetting, this
choice probably won't matter too much.

> I also suspect that fixed point representation would also speed up
> interpolation quite lot, since currently the interpolation weights are
> double, and many expensive byte -> double conversions will be done.

Will 32 bit fixed point be sufficient for HDR images, or will you only
switch to fixed point for lesser bit depths?

Good luck,

JD