基于反向光线跟踪技术和感兴趣区域的超多视点图生成加速算法
首发时间:2020-02-14
摘要:对于包含运动物体的复杂场景而言,使用计算机生成用于三维显示的超多视点图的效率往往较低。为了解决该问题,本文提出了一种基于反向光线跟踪技术以及感兴趣区域的生成加速算法,可以在保证显示质量的前提下对超多视点图的生成过程进行加速。在每一帧中,内容变化的部分将被检测并标记为感兴趣区域,只有该部分的像素将被重新渲染。实验结果表明,内容生成过程的速率可被提升至原来的2~9倍。此外,在分辨率为3840x2160的情况下,输出帧率由3.5帧/秒提升至超过30帧/秒,满足了实时交互的需求。
For information in English, please click here
An accelerating algorithm for generation of super-multiview images based on the region of interest and backward ray-tracing technique
Abstract:To overcome the limitation that computer generated super-multiview (SMV) images for three-dimensional (3D) display is not efficient enough for complicated scenes with the moving object, an accelerating algorithm is presented. The generating process can be accelerated based on the backward ray-tracing technique and the region of interest (ROI) while the image quality is maintained. For each frame, the changed part of the scene is detected and marked as the ROI, and only the pixels located in the ROI are re-rendered. Experimental result shows that the generating process can be accelerated to be 2~9 times faster. In addition, the output frame rate is improved from 3.5 frames per second (FPS) to over 30 FPS with a resolution of 3840x2160 for the complex application situation, which meets the real-time demand.
Keywords: auto-stereoscopic display, ray-tracing, super-multiview image, three-dimensional display,
引用
No.****
动态公开评议
共计0人参与
勘误表
基于反向光线跟踪技术和感兴趣区域的超多视点图生成加速算法
评论
全部评论0/1000