Advanced Flex / Pixel Bender interactions

Question

Hello,
I've been looking through the documentation on both Pixel Bender and Flex, and I'm trying to figure out if this even possible:
Basically I start out with floating point image data in a ByteArray and I'd like to run a Pixel Bender filter or shader job on the data in floating point, then convert it to 8-bit / channel RGBA on the output.
I know I can start with a 8-bit BitmapData and use a Pixel Bender shader to produce a new 8-bit BitmapData. Or I can start with a floating point ByteArray, use ShaderJob to compute a new floating point ByteArray.  Can I use ByteArray input and output to BitmapData?  Put another way, do the input and output classes/dimensions of a shader job or pixel bender filter have to match?  I'd like to take advantage of the fact that when using a BitmapData object, Pixel Bender (or Flash), converts the floating point Pixel Bender data to 8-bit for you automatically.  It would also save me the trouble of unpacking a linear list of floats into a 2-dimensional image with setPixel32() calls.
Thanks all.

Answer 1

Hi David -
Thanks for those examples. The audio visualizer is pretty cool, but looks like the author witheld the source code for us to look at .
I did find another blog posting that's helpful too - converting a BitmapData object into a Vector of RGBA floats with Pixel Bender:
http://webr3.org/blog/general/pixelbender-is-more-useful-than-i-assumed/comment-page-1/
I'll see if I can figure out going backwards.
I'm curious if anyone could shed light on what's going on with memory under the covers. I found another blog post that treats a Vector as if it's a 2-dimensional object like an image:
http://blog.robskelly.com/2009/02/using-pixel-bender-for-math-in-flashflex/
This raises questions about what do the width and height input/output parameters actually control if you can treat a flat Vector as an object with width and height?  Does anyone know? Does that only work with Vectors? What about ByteArrays?  Does this basically mean that width/height are only convenience parameters and you could put any 2 numbers that multiply to the size of your data?   How does this affect the input type inside the shader (image1 vs image4)?  Why does Rob Skelly's example above even work at all?
I'll hack away at trying to get stuff working, and re-post if I figure something out.