摘要
随着三维视频资源需求的增长，二维转三维（2D-3D）转换技术成为三维（3D）图像处理技术研究的焦点．由于2D视频中没有包含深度信息，2D-3D转换技术的关键在于2D视频中深度信息的提取．提出一种基于最小生成树的二维图像深度分配方法．利用最小生成树的方法对图像进行分割，然后先验理论对分割区域进行深度分配以产生合适的深度图像．最小生成树的方法不需要设置较多参数，对于不同场景的图像有较强的自适应性，而且最小生成树算法具有算法简单，运行速度快的优势．
With the development of demand for three-dimensional video resources, three-dimensional （2D-3D） conversion technology become the the focus of three-dimensional image （3D） processing technology research. However, 2D video does not contain depth information.Therefore, the key technique of 2D-3D conversion is the extraction of the depth information in 2D video. This paper presents a depth assignment methods for two-dimensional map based on minimum spanning tree（MST）. The methods use the an approach based minimum spanning tree to do image segmentation, and then assign a depth value to every region according to a prior theory, to produce a right depth map. Minimum Spanning Tree method does not require setting many parameters, and has a strong adaptability for images of different scenes, moreover, minimum spanning tree algorithm is simple,and has a advantage of running speed.
作者
高鹏
魏姗姗
石晓龙
胡天宇
沈涛
陈从周
GAO Peng;WEI Shan-shan;SHI Xiao-long;HU Tian-yu;SHEN Tao;CHEN Cong-zhou(School of Automation, HUST Wuhan 430074, China;MOE Key Laboratory of Image Processing and Intelligence Control, Wuhan 430074, China)
出处
《数学的实践与认识》
北大核心
2018年第9期163-167,共5页
Mathematics in Practice and Theory
基金
国家自然科学基金(61572213)
湖北省自然科学基金(2014CFB730)
