Re: Bitmap code - 3rd alpha version

[Stefan Boehringer]

Hi Pete,

hm..., there are alternatives in between your two implementations. You may  
choose a fixed set of points over which to take the average. E.g. you choose a  
rectangle which is to be collapsed into a single point. Then you may choose,  
say 9 points, equally distributed in that rectangle and average over them. Just 
 
a thought.

Stefan

Pete French wrote:
> O.K., a 3rd try. This one includes optimisations for the visibleon
> screen rectangle so non visibleparts are not calculated. This will
> speed up code somewhat.
>
> Also, ater all the "its slow" comment, I added the flag to disable the
> nice minification. By default the code now compiles to just choose the
> nearest point when the destination area is smaller than the source. This
> is a lot faster (or rather the speed is relative to the size of the
> destination as opposed to the source). I got a speedup of
> about 15 times rendering a 1000x1000 tiff into a 150x200 view.
>
> It does look ugly though - o.k. on photographs, but anything with fine
> detail or grids on it gives unpleasent aliasing effects. As I work with
> maps this is a bit horrid to me.
>
> If you want to re-enable the old behaviour then line 1700 in XGGState.m
> sets the fast minimisation flag to YES - change it to NO and the old
> behaviour comes back. Somehow I want this to be selectable at runtime for
> particular images, but for now live with it as a compile option.
>
> cheers,
>
> -bat. [patch relative to current CVS code]