基于单张图像的彩色三维重建技术研究  

Research on color 3D reconstruction technology based on single image

作　　者：姜慧 任小强 蒋玉香 JIANG Hui;REN Xiaoqiang;JIANG Yuxiang(Department of Information Engineering,Hope College,Southwest Jiaotong University,Chengdu,Sichuan,China 610400)

机构地区：[1]西南交通大学希望学院,四川成都610400

出　　处：《深圳信息职业技术学院学报》2024年第6期44-51,共8页Journal of Shenzhen Institute of Information Technology

基　　金：国家自然科学基金项目(项目编号:62076200);四川省彝族文化研究中心项目(项目编号:YZWH2347)。

摘　　要：基于单张图像的彩色三维重建在虚拟现实、增强现实、电子商务产品的展示与推广等领域具有广阔的应用前景。目前的研究主要集中于从单张图像中重建物体的形状信息,忽略了物体的表面颜色重建。提出新方法对物体形状进行彩色三维重建,首先基于“编码器-解码器-细化器”的网络结构对形状重建;其次采用相同的编码器结构重建颜色,使用解码器生成3通道的颜色体素,将递归颜色体素直接从输入的图像中学习并获取颜色信息,通过融合从图像中采样得到的代表性颜色体素,再混合颜色体素;最后,将颜色重建分支和形状重建分支得到的结果进行融合,得到彩色的三维对象。Color 3D reconstruction based on a single image holds vast potential in fields such as virtual reality,augmented reality,and the display and promotion of e-commerce products.Current research primarily focuses on reconstructing the shape information of objects from a single image,neglecting the reconstruction of object surface colors.This research proposes a new method for color three-dimensional(3D)reconstruction of object shapes.First,the shape is reconstructed based on the network structure of"encoder-decoder-refiner"';secondly,the same encoder structure is used to reconstruct colors,and the decoder is used to generate three-channel color voxels.The recursive color voxels are directly learned from the input image and color information is obtained.By fusing representative color voxels sampled from the image,mixed color voxels are obtained;Finally,the results from the color reconstruction branch and the shape reconstruction branch are fused to obtain a colored 3D object.

关 键 词：单视图 深度学习 形状重建 颜色重建 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术] TP18[自动化与计算机技术—计算机科学与技术]