> Instead, I'd advise you look at a channel-based representation: break  
> up RGBA into 4 larger buffers and pass those to your  
> "doSomethingWithImage" function. 

Mmm, I agree, however I would add that the ideal data format depends on the
operation being performed and the data source format. Its not a simple problem.

My personal interest is in video and so I am hot on image processing code
which makes no assumptions about how many frames there are - it could even
be could be infinite (resample /dev/tv | page). Also in video many image
processing algrorithms require access to temporarly adjacent frames (or fields).

I put down some thoughts about what I would like to do (but will probably
never get time) here: /n/sources/contrib/steve/Blog

Please don't think I am telling what you should do, just point you at some
random thoughts which will probably raise more questions than they solve ☺

-Steve

Reply via email to