Beyond the Frame: Perceptually Based Rendering with Distributed Graphics Systems

0135943 David Luebke哥伦比亚大学计算机界越来越认识到理解和利用人类感知系统的重要性。事实上，一些研究人员已经提出了通过利用人类视觉的局限性来加速渲染的技术。 然而，这些研究主要集中在真实感绘制算法上，如光线和路径跟踪，这些算法需要几秒钟或几分钟的时间才能完成绘制，而感知驱动的交互式绘制算法却很少，因为每幅图像都必须在几毫秒内完成。3-D模型复杂性的增长似乎超过了我们渲染它们的能力。研究感知驱动的交互式渲染加速方法的一个障碍是标准图形硬件的兼容性。经过是的的专门化和完善的渲染流水线现在在一组狭窄的渲染任务中非常快，但研究人员不能篡改传统的渲染基础。 本文提出了一种基于集群的感知渲染方法，该方法主要研究了几种感知渲染技术在传统的面向对象的交互式渲染模式下无法实现的问题。

0135943David LuebkeUniversity of VirginiaThe computer g aphics community increasingly recognizes the importance of understanding and leveraging the human pe ceptual system.Indeed,several researchers have proposed techniques to accelerate rendering by exploiting the limitations of human vision. However,this research has focused on realistic rendering algorithms such as ray and path tracing, which take seconds or minutes to render.Ve y little work has been done on perceptually d iven interactive rendering,whe e each image must be completed in milliseconds.Yet a dire need exists for techniques to accelerate interactive rendering of ever-la ger datasets,fo despitetremendous strides in graphics hardware,the growth in complexity of 3-D models appears to beoutstripping our ability to render them.One barrier to research on perceptually motivated interactive rendering acceleration methods has been the inflexibility of standard graphics hardware.The rendering pipeline, specialized and perfected over many yea s,is now extremely fast at a narrow set of rendering tasks,but researchers cannot tamper with the traditional polygon-rendering underpinnings. We propose to circumvent this barrier by using software rendering,distributed across a cluster to achieve interactive ates on complex scenes.This SGER focuses on building the equipment necessa y to analyze several perceptual rendering techniques impossible under the traditional polygon-oriented interactive rendering paradigm.