基于贝尔格式原始图像的超分辨率三维几何建模技术

项目名称： 基于贝尔格式原始图像的超分辨率三维几何建模技术

项目编号： No.61201376

项目类型： 青年科学基金项目

立项/批准年度： 2013

项目学科： 电子学与信息系统

项目作者： 周祚峰

作者单位： 中国科学院西安光学精密机械研究所

项目金额： 24万元

中文摘要： 三维几何建模技术一直是计算机视觉和图像处理等领域的研究热点。本项目将基于图像的三维几何建模技术、Bayer格式原始图像彩色插值技术与超分辨率图像重构技术相结合，研究基于Bayer格式原始图像的超分辨率三维几何建模技术。首先通过偏微分方程和张量分析理论，研究基于单帧Bayer格式原始图像的超分辨率三维几何建模算法；其次研究建立三维空间中相机标定和二维空间中运动估计的联系，并以此作为结合点，利用空间自适应P范数正则化理论，研究基于多帧Bayer格式原始图像的超分辨率三维建模算法。本项目的最终目标是利用不同视角下物体的低分辨率Bayer格式原始图像通过联合优化得到对应视角的高分辨率彩色图像和高分辨率三维模型，为后续的目标识别和虚拟现实等实际应用奠定良好的基础。

中文关键词： 三维建模；单帧超分辨率图像重构；多帧超分辨率图像重构；；

英文摘要： The 3D geometrical modeling technique is a research hot point in computer vision and image processing applications. The proposal will study the super resolutino 3D geometrical modeling technique based on Bayer pattern raw images by combining the traditional 3D geometrical modeling, image demosaicking and the super resolution image reconstruction method. First, the super resolution 3D geometrical modeling based on single-frameBayer pattern raw image is investegated by joint partial differential equation and tensor analysis theory. Then, we will establish the relationship between the camera calibration in the 3D space and the motion estimation in the 2D space. Based on this relationship, the multi-frame Bayer pattern raw images super resolution 3D geometrical modeling is investigated via the spatially adaptive P-norm regularization method. Given the low resolution Bayer pattern raw images under different view, The taks of our proposal is to get the corresponding high resolution color images and high resolution 3D geometrical model simultaneously via joint optimization, which can facilitate in some pratical applications, such as target recognition and vitual reality.

英文关键词： 3D geometrical modeling；single image super resolution；multi-frame super resolution；；