Re: Source code or algorithm for these effects??

"eminemence" <eminemence@xxxxxxxxx> wrote in message
news:1130335730.024558.9930@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
> So for implementing a transition effect rather the blurring effect has
> quite some cost in terms of cpu and memory.

The cost per interpolation of two images is smaller than blurring
a single image. The issue is how many interpolations you want.

> I tried blurring and this time its working and after applying the
> filter once it
> did not blur properly.I got some random colors which can be attributed
> to the division which may not be approximating to the correct color.

Your blur is applied to RGB colors. You might want to consider
some other color space. I am not an expert on those, so other
folks need to tell you which color space to use.

> Also just applying the filter once took about 2 seconds,so for the
> transition effect
> it will surely take more time and processing power.
> Would you support me in this claim?
> So should I conclude that its a very bad idea to implement blurring as
> a transition effect on ARM processors with no floating point units?

I have implemented transitions on an ARM processor with no
floating-point unit. The performance was acceptable. The
number of intermediate steps in the transition is yours to choose.
The larger the number, the more cycles you need.

One thing you should be aware of is that the graphics system
on embedded devices typically perform miserably when you
set one pixel at a time. What you should do is draw to a
chunk of system memory, and then BLT that to the graphics
system--double buffering, so to speak. I know this greatly
improved performance on the ARM processor and graphics
system I was working on.

--
Dave Eberly
http://www.geometrictools.com


.