Help Developing Better Velocity Blur COP

[AndrewJBeard]

Would anyone like to help me develop a proper vector-based motion blur cop? I always prefer to defer my motion blur pass to post if I can, and I've found it difficult to work with the Velocity Blur COP and would rather not have to pipe my renders through a third-party solution (such as RSMB).

Because of its VEX origins (I believe), the Velocity Blur COP does not smear the current pixel value along the velocity vector and instead looks up surrounding pixel values and their velocity vectors to reverse-engineer the effective blur at that pixel (VEX can only shade the point/pixel it is evaluating). The difference is subtle, but the current implementation creates artefacts at object boundaries, and is terribly slow.

I reckon that a better implementation could be written in C++ - and I'd be willing to have a go at it myself, but I don't have access to the appropriate developer tools (VS2005?) and probably don't know enough about programming in C++. If anyone feels they can help me out with writing a better, faster, velocity blur COP then I'd welcome the support. I'd be more than willing to work on the pseudo-code of the blur operation itself.