Still trying different interpolation methods, such as Inverse Distance, RBF, Spline etc.
Believe me, in DD’s STORM or Sony imageworks’s SplatQ or 5x5 would not use ANY ET’s technology, just depended on rich experience and smart tricks heavily.
OpenMP enabled, the more CPU core the more faster. CUDA-based solution will be involved soon.
依旧在尝试不同的差值方法，包括反向距离权重、RBF、样条等等。
相信我，无论是DD的STORM还是SONY的SplatQ或者是5x5，它们都不可能是外星科技的产物，它们肯定是丰富制作经验与技巧结合的结果。
目前的是基于OpenMP的多线程（主要是我懒得专门写MT），自动检测多少个核心自动加速。CUDA将在性能测试后引入。