On Jan 31, 2008, at 16:12 , Juan Vuletich wrote:

> Bert Freudenberg wrote:
>>
>> The way I would do your demo in OpenGL would be that every morph  
>> renders to its own off-screen surface, and when compositing, it  
>> would apply the non-linear transformation. This is now a standard  
>> implementation technique, except it usually is not used for  
>> projective geometry but for fancy effects (like the Genie effect  
>> when minimizing Mac OS X windows).
>>
>> Mind I did not understand what you said in the talk, just  
>> commenting on the visuals ...
>>
>> - Bert -
>>
>>
> Yes. But you would be doing only a part of it in OpenGL. The  
> geometric deformations would still be needed, but working only on  
> bitmaps. The quality would not be as good (for things like  
> CurveMorphs) because the off-screen surface would have some  
> specific resolution, that could be too coarse for those areas that  
> are enlarged by the non-linear transformation. Think for example on  
> a graph with logarithmic X. Pixels in the off-screen surface would  
> be too big at the left of the graph and too small at the right.

I am beginning to understand your point :) Yes, having that power in  
the base system would be cool. I still think it can be implemented on  
latest-gen OpenGL hardware (which can do the non-linear transform and  
adaptively tesselate curves to pixel resolution) but that then would  
be just an optimization.

- Bert -