项目名称: 面向运动模糊的场景组织方法的研究
项目编号: No.61300084
项目类型: 青年科学基金项目
立项/批准年度: 2014
项目学科: 自动化技术、计算机技术
项目作者: 杨鑫
作者单位: 大连理工大学
项目金额: 23万元
中文摘要: 运动模糊是一种在照相机摄影过程中产生的影像效果,能给场景绘制带来极大的视觉真实性。然而,由于额外的采样计算和面片使用,实现准确的运动模糊效果增加了场景组织的复杂度,这些额外的复杂性可能引起低效甚至错误的场景层次结构,使运动模糊效果的快速绘制变得更加困难。针对上述问题,本项目主要研究如何根据运动模糊特征对场景组织的影响,从对象划分、空间划分、细节划分三个角度设计合理的加速结构来提高场景组织的有效性,降低运动模糊效果的计算复杂度,最终提升绘制的性能。首先,研究面向运动模糊效果绘制的运动行为分类方法,并基于时空维度信息的转化关系,有效处理不规则运动和破坏层次结构的面片;其次,提出基于运动模糊强度的多层次细节表示方法;此外,能够使访存不友好的遍历操作尽可能转化为计算密集型的相交操作,以优化存储访问。本项目预期在图形学理论的原始创新方面提供一种有价值的探索,并最终推动虚拟现实、电影游戏等产业的发展。
中文关键词: 运动模糊;场景组织;采样技术;加速结构;图形绘制
英文摘要: Motion blur occurs in photography whenever the camera or objects move while the shutter is open, and brings scene rendering a realistic visual effect. However, implementing accurate motion blur effects significantly increases scene organizing complexity due to either extra geometry, sampling, or both, likely causes inefficient even wrong scene hierarchy. This additional complexity makes real-time motion-blur computation difficult. In this proposal, we investigate the interrelation between acceleration structure and motion blur from object partition,space partition and detail subdivision. We intend to design efficient acceleration structures to improve the efficiency of scene organizing, reduce computation complexity, finally increase rendering performance. First, we present movement behavior classification for motion blur, intend to solve irregular primitives and some primitives worsening scene hierarchy using information transformation between time and space dimension.Secondly, we propose a novel blur-dependent level-of-detail scheme. Besides, we devise to improve memory access, by transferring traversal operation with incoherent memory access to intersection operation with intensive computation. We expect this research can provide a valuable exploration for graphics theory innovation, and contribute the develo
英文关键词: Motion Blur;Scene Organizing;Sampling Technology;Acceleration Structure;Rendering