The Research And Development Of The Real-time Rendering Framwork And Features In The Virtual Home Design System

This paper introduces and analyses the real-time rendering framework and features of the virtual home design system called EasyHome. Firstly, based on the system’s particular needs, we design and implement the real-time rendering framework, which can be divided into two stages, including the application stage and rendering stage. In the application stage, we fulfill the Cell&Portal algorithm which is suitable for indoor scene management to do visibility test efficiently. In the implementation details, we construct the Cell&Portal graph in dependent on the walls, windows and doors’information in real time. In the rendering stage, we uses the recently popular real-time rendering framework called deferred lighting to do pixel lighting efficiently. And in order to improve the naive framework’s performance, we add some optimizations, including compressing the geometry and lighting buffers, using light volume algorithm to identify the lit pixel efficiently and accurately and modifying the rendering equation to meet the deferred lighting framework. With these optimizations, our system eliminates the multi render-target video card limitation and supports multi lights meanwhile. Secondly, we introduce the soft shadow generation solution in details. In the indoor scene, we use the Variance Shadow Map algorithm to produce soft shadow. In order to reduce the light bleeding artifact, we modify the Chebyshev inequality with a negative tail and combine VSM with another statistics based shadow map algorithm called Exponential Shadow Map. In the outdoor scene, we use the Light Space Perspective Shadow Map algorithm to do Anti-Aliasing and combined with Percentage-Closer Soft Shadows to generate soft shadows. Lastly, we introduce one global illumination algorithm called Screen Space Ambient Occlusion. In the implementation details, we use the mixed resolution rendering technique to produce and blur the AO value efficiently, and combine filter techniques such as Gaussian filter and bilinear filter with bilateral filter during the blur and up-sampling pass in order to preserve the edges.