Improved the core model simplification algorithm #28
Conversation
…takes into consideration the discrete curvature to the quadric error metrics in order to retain detailed features in model simplifcation. Based on a paper by Li Yao and Shihui Huang.
|
Hi @bawar9, Thank you for your contribution! |
There was a problem hiding this comment.
Great job on this!
I have tested it, and it does indeed improve the simplification in some cases.
There are a few performance boosts that you can do (as I comment on below), and a few code style fixes.
I hope that you don't mind, but I would like to try to optimize this even further after it has been merged.
…riangle utility functions to take reference type parameters instead of value type structs.
|
Hi, thank you for pointing out the issues and sorry for being sloppy. Only now did I realize that the "Triangle" and "Vertex" are structs and not classes, If I knew that earlier I would've passed them by reference to all the new methods and please do let me know if you find any other optimization hotfix. I have also removed the normal calculation method and have made all the other requested changed as well. |
Hi, these changes are based on a research paper that augments the original paper by Dr Garland. You can read it here: https://www.hindawi.com/journals/mpe/2015/428917/. This also solves the issue listed here to some extent #19. See for yourself below.
BEFORE MY CHANGES
LOD Level (Quality 0.55 - Total Triangles after reduction 1208)ca
AFTER MY CHANGES
LOD Level (Quality 0.55 - Total Triangles after reduction 1208)ca
Please note that this is not meant to exclusively solve the boat problem originally reported here
#19. In fact the issue with the boat model is still there and something tells me that it's related to the vertex linking.