Last week a succeeded in getting the half-edge collapse operator working, I think I need to put more time into the error metrics now so that I can get it to produce higher quality output and hopefully extend it to perform full-edge collapses. I am currently looking into the maximum supporting plane distance, introduced by Ronfard and Rossignac in 1996, and error-quadrics which is an extension of maximum supporting plane distance developed by Garland and Heckbert in 1997. I hope to have at least one of these methods implemented this week.
I have integrated Nvidia PerfHUD into my viewing application so that I can get profiling data, I decided to use it instead of the Direct3D9 query functionallity do to the simplicity of integration and the automatic visualisation. Because both my home computer and the university computers run on Nvidia graphics cards I felt that being platform dependent wasnt an important factor in my desision.
Tests on my home computer using a scene comprising of 25 very high polygon models have resulted in a frame rate of approximatly 28fps. Substituting the output from my half-edge collapse increased the frame rate to 39fps with only marginal visable loss of detail in the nearest model, this will because I have yet to implement any switching methods so it was just using the lower resolution model.
Subscribe to:
Post Comments (Atom)
No comments:
Post a Comment