2维至3维图像/视频转换的深度图提取方法综述

Depth map extraction methods in 2D-3D image/video conversion

目的深度图提取是计算机视觉领域的研究热点。随着3D显示设备的普及,2D-3D图像/视频转换的深度图提取研究受到越来越多国内外学者的关注。为此回顾深度图提取研究历程,并对已有成果进行分类、概括和评述。方法由于深度图提取方法的实现主要依赖于深度线索,不同方法存在人机交互程度上的差异。采用基于深度线索和基于人机交互程度的两种分类方法进行归纳评述。结果 根据深度线索的不同,将深度图提取方法分为基于单目线索的方法、基于双目线索的方法和基于混合线索的3类方法。然后从人机交互的角度,将深度图提取方法分为人工法、半自动法和全自动法。介绍了这些方法的基本思想,比较归纳不同方法的优点与不足。最后,阐述了近年来热门的机器学习方法在深度图提取的应用。结论对深度图提取研究进行简要的总结和展望。指出深度图提取研究具有从研究热点中挖掘创新思路、引入新的深度线索等发展趋势。

Objective Depth map is a hotspot in computer vision research field.With the development and promotion of 3D display equipment,depth map extraction methods have attracted considerable attention.This study reviews the development trend of depth map extraction and summarizes the existing methods.Method Classifications are adopted on the basis of depth cues and human-computer interaction degree.Result The existing methods are grouped into three categories,namely,monocular,binocular,and multiple depth cue-based types.These methods are then classified into three schemes based on different human-computer interactions,namely,manual,semi-automatic,and automatic.This study focuses on the basic principles of these methods and emphasizes their advantages and limitations.A detailed analysis is also performed based on the application and development of machine learning methods for depth extraction.Conclusion Future developments in depth extraction,including the adoption of new methods and the introduction of new depth cues,are discussed.