摘 要
绚烂多彩的自然界光照渲染是计算机图形学的永恒追求。传统的基于几何模型的渲染方法无法很好的处理复杂动态的场景，于是基于图像的渲染方法被提出，尤其是基于图像的光照技术，极大的增强了渲染场景的真实感。
传统的 IBL 技术使用预计算辐照度贴图的方法计算漫反射光照，高效快速。但由于镜面反射光照的复杂性，传统的 IBL 技术无法应用预计算等方式有效快速的处理，导致传统 IBL 技术渲染时间较多，一般多用于离线渲染。本文在传统的 IBL 工作基础之上， 改进了间接光照中镜面反射的计算，提高了渲染质量，极大的提升了渲染效率，达到了实时渲染的要求。
本文使用分割求和近似方法，拆解了镜面反射的积分计算，分别预计算处理镜面反射光照与 BRDF 部分，保证了镜面反射的实时计算。除此之外，由于镜面反射并非均匀分布，故均匀采样不适用于此，效率过低。基于镜面反射光照分布依赖于表面粗糙度， 多集中于高光波瓣这一观察，本文使用 GGX 重要性采样技术处理镜面反射光照的卷积运算，定位偏移，使得采样向量多集中在高光波瓣范围内，加快收敛速度。
实验证明本文算法在渲染质量和渲染效率上均远高于传统 IBL 技术，降低了镜面反射的计算开销，提升了渲染质量和渲染效率，能够快速进行实时渲染。
关键词：基于图像的光照；实时渲染；重要性采样；分割近似求和
Abstract
Abstract
Colorful natural lighting rendering is the eternal pursuit of computer graphics.The tradi- tional rendering method based on geometric model can not deal with the complex dynamic scene well, so the image based rendering, especially the image-based lighting technique, is proposed to greatly enhance the realism of the rendering scene.
The traditional IBL technology uses the method of pre-calculated irradiance map to cal- culate diffuse lighting, which is efficient and fast. However, due to the complexity of specular reflection lighting, traditional IBL technology cannot apply pre-calculation and other methods to effectively and quickly process, resulting in a long rendering time of traditional IBL tech- nology, which is generally used for offline rendering. Based on the traditional IBL work, this paper improves the calculation of specular reflection in indirect lighting, improves the render- ing quality, greatly improves the rendering efficiency, and meets the requirements of real-time rendering.
In this paper, the split-sum approximation method is used to disassemble the integral cal- culation of the specular reflection, and separately pre-calculate the lighting and the BRDF part to ensure the real-time calculation of the specular reflection. In addition, since the specular re- flection is not uniformly distributed, uniform sampling is not suitable for this, and the efficiency is too low. Based on the observation that the specular reflection light distribution depends on the surface roughness and mostly focuses on the specular lobe, our paper uses GGX importance sampling technology to process the convolution operation of the specular reflection light, and through the positioning offset, the sampling vector is mostly concentrated in the specular lobe, to speed up the convergence process.
Experiments have proved that the algorithm in this paper is much higher than the tradi- tional IBL technology in terms of rendering quality and rendering efficiency, which reduces the computational cost of specular reflection, improves rendering quality and rendering efficiency, and can quickly perform real-time rendering.
Keywords: 关键词的英文翻译要准确
