摘要
在计算机图形学中全局光照的效果直接影响画面的真实性,使用传统的光线跟踪技术计算复杂,难以实时。所以在游戏等实时应用中一般选用环境光遮蔽(AO)技术模拟全局光照效果。屏幕空间环境光遮蔽(SSAO)是实际应用较多的一种AO算法,该算法可以在实时运行的条件下较为逼真的模拟全局光照的渲染效果。针对目前SSAO算法在采样点选择、纹理采样和平滑滤波3个方面的缺陷,提出相应的改进方法,改善原有算法的性能和渲染效果,使其更适用于三维游戏中的实时游戏场景渲染。
In the process of face attribute recognition, there are several methods at present, face attribute recognition based on Gabor wavelet transform, face attribute recognition based on SIFT and face attribute recognition based on differential texture features. There are many problems in the traditional methods, such as the characteristics of the selection needs human intervention, and the characteristic of the choice are not necessarily able to meet expectations. Adopts deep convolutional neural network(DCNN) based on the supervised method, constructs a multilayered convolution neural network, neural network obtained by convolution convolution activation depth features, this method uses the Celeb A database training, after using the JAFFE face database for testing, and achieves good results.
引文
[1]Landis H.Production-Ready Global Illumination[C].New York:Proceeding of the International Conference on Computer Graphics and Interactive
[2]Morgan Mc Guire,Michael Mara,David.Scalable Ambient Obscurance[J].High Performance Graphics,2012.
[3]Lance Williams.Pyramidal Parametrics[J].Computer Graphics,1983.
[4]Morgan Mc Guire,Brian Osman,Michael Bukowski,Padraic Hennessy.The Alchemy Screen-Space Ambient Obscurance Algorithm[J].High Performance Graphics,2011.
[5]Louis Bavoil,Miguel Sainz,Rouslan Dimitrov.Image-Space Horizon-Based Ambient Occlusion[Z].NVIDIA Corporation,2009.
[6]Ville Timonen.Screen-Space Far-Field Ambient Obscurance[J].High Performace Graphics,2013.
[7]Thai-Duong Hoang,Kok-Lim Low.Efficient Screen-Space Approach to High-Quality Multi-Scale Ambient Occlusion[J].The Visual Computer,2011.