On Sunday 08 March 2009 14:38:51 Stefan Krastanov wrote:
> Hi,
>
> First about those a, b, c coefficients - yes they were dynamically
> calculated for every iteration, but they worked quite crappy - so here is
> another idea that is extremely simple, already implemented and working
> quite well.
>
> So what's the idea. We don't need to know the relation between exposure and
> yavg ("linear", "square" or "linear with correction"). It's enough to know
> that exposure goes up when yavg goes up. So if yavg is too low we add a
> little bit to exposure and if yavg is too high we substract a little bit
> from exposure. To get rid of the oscillations we keep track of the
> direction of the last two chainges - if they are in opposite directions we
> make the step smaller, if not we make it bigger. It's working with me, I
> tested it all ready. I will be very happy to give you the sorce if you
> think it's usefull, but I don't know how "patch" works. What should I do.
>
> There is a question left - when changing the exposure it seems that for my
> chip the maximum is 83 (decimal) and not 0xFF. Is this normal.Ok, but how to calculate this step? I tried this approach before, but it takes up to 3-5 seconds to stabilize :(
signature.asc
Description: This is a digitally signed message part.
