[PD] Recommendations for combining many images (256+)

B. Bogart ben at ekran.org
Sat Sep 20 18:05:12 CEST 2008


Hi Chris,

I have used SB before for this, but in this case I need more flexibility
(and the images should not build up over time but be combined from the
first frame.

Seems like I may have to delve into some C++ to do the math manually.
Will that be too slow though? iterating over the data in a pix_buffer
should be fast right?

If anyone has some sample code along these lines please let me know.

B. Bogart

chris clepper wrote:
> Have you tried single buffer rendering?  If you don't need 30fps
> performance you could layer the images with opacity using only one
> pix_texture object and very little GPU memory.
> 
> On Fri, Sep 19, 2008 at 3:59 PM, B. Bogart <ben at ekran.org
> <mailto:ben at ekran.org>> wrote:
> 
>     Hey all,
> 
>     Any recommendations as to the best way to combine many images?
> 
>     What I'm after is a kind of long exposure where each image could be
>     weighted differently in order to give it more emphasis.
> 
>     I did a quick test in Gem layering 100 images on 100 rects on top of one
>     and other with 0.01 opacity.
> 
>     Problem it uses too much CPU (seems to be the GPU lagging). I can render
>     all 100 easily on the same machine if they are not totally overlapping.
> 
>     Would gridflow be able to combine 256 640x480 images?
> 
>     CPU solutions (rather than GPU) are best, as I'm hoping for 640x480
>     resolution and many more images than will fit in graphics mem.
> 
>     Would something like this work:
> 
>     Every channel of every pixel of every image divided by the number of
>     images (256) multiplied by a weight, where all the results are summed?
> 
>     All the images would be loaded into a pix_buffer.
> 
>     What is the best way to accomplish this?
> 
>     Thanks,
> 
>     .b.
> 
> 
>     _______________________________________________
>     Pd-list at iem.at <mailto:Pd-list at iem.at> mailing list
>     UNSUBSCRIBE and account-management ->
>     http://lists.puredata.info/listinfo/pd-list
> 
> 





More information about the Pd-list mailing list