光场深度估计方法的对比研究

发布时间：2018-01-16 14:03

  本文关键词：光场深度估计方法的对比研究　出处：《模式识别与人工智能》2016年09期 　论文类型：期刊论文

  更多相关文章： 光场 深度估计 光照 极平面图像 重聚焦

【摘要】：为了更有效地利用光场信息实现场景深度的精确估计,文中回顾并深入探讨光场的深度估计问题.通过阐述光场基本理论,将光场深度估计归纳为基于极平面图像、多视角图像及重聚焦的3种方法.在合成数据集上,对比光照变化对不同算法性能的影响,并构建一个更全面且具有挑战性的光场数据集.在该数据集、光场标准数据集及Lytro Dataset上,定性及定量分析不同复杂场景对算法性能的影响,进一步指出该领域的研究方向.
[Abstract]:In order to use the light field information more effectively to realize the accurate depth estimation of the scene, the depth estimation problem of the light field is reviewed and discussed in this paper, and the basic theory of the light field is expounded. The depth estimation of light field is divided into three methods based on polar plane image, multi-view image and refocusing. The effects of illumination on the performance of different algorithms are compared on the composite dataset. A more comprehensive and challenging light field data set is constructed on the data set, optical field standard data set and Lytro Dataset. The effect of different complex scenes on the performance of the algorithm is analyzed qualitatively and quantitatively, and the research direction in this field is pointed out.
【作者单位】： 合肥工业大学计算机与信息学院;
【基金】：国家自然科学基金项目(No.61403116,61271121) 中国博士后基金项目(No.2014M560507) 中央高校基本科研业务费专项资金资助~~
【分类号】：TP391.41
【正文快照】： Supported by National Natural Science Foundation of China(No.61403116,61271121),China Postdoctoral Science Foundation(No.2014M560507),Fundamental Research Funds for the Central Universities深度感知(Depth Perception)是指人眼对物体远近距离的感觉.人眼的深度，

本文编号：1433409