项目名称: 高质量本征图像求解技术及其应用研究
项目编号: No.61272359
项目类型: 面上项目
立项/批准年度: 2013
项目学科: 自动化技术、计算机技术
项目作者: 沈建冰
作者单位: 北京理工大学
项目金额: 81万元
中文摘要: 本征图像求解主要研究从图像中提取亮度和反照率信息构成亮度本征图和反照率本征图。物体表面所呈现的颜色是由物体的材质和光源的颜色等本征属性决定的,而反照率和亮度信息是两种最重要的本征属性。恢复单幅图像的本征图仍然是计算机图形学与计算机视觉领域的研究难点之一。因此研究高质量本征图像的求解技术及其应用,不仅具有重要的理论意义,而且在计算摄影学和真实感图形学绘制等领域具有重要的应用价值。本项目具体包括基于单光源/多光源混色双色反射模型(MHDRM)的高质量本征图像求解方法、基于子带结构和稀疏表达的自动本征图像计算方法、基于隐含狄利克雷分布模型(LDA)学习过程的高质量本征图像求解算法、基于本征图像的实时图像着色与二次着色方法、基于本征图像/视频的实时材质编辑方法与纹理替换算法。项目最终将实现一个基于GPU 技术的实时高质量本征图像/视频求解系统,并将相关技术应用于真实感图形绘制和影视特效制作等
中文关键词: 本征图像;能量优化;纹理替换;阴影检测;
英文摘要: Intrinsic images recovery algorithm usually aims at separating the illumination and reflectance components from an input photograph. The observed color on an object surface is influenced by many intrinsic factors, including the material of the object and the colors of the light sources, while the two of the most important characteristics are the shading and reflectance in the scene. Therefore, it is still an open challenge on how to recover the high quality intrinsic images, and has the fundamental applications in the areas of computational photograph and realistic graphics rendering. The main research topics include the Single- illuminant and Multi-illuminant Hybrid Dichromatic Reflection Mode(MHDRM) based intrinsic images, the subband architecture and sparse representation based intrinsic images recovery approach, the Latent Dirichlet Allocation (LDA) learning based intrinsic images, the colorization and recolorization for images and videos using the intrinsic images, and the intrinsic images based image/video retexturing techniques. Finally, we develop a GPU-based real time system for recovering the high quality intrinsic images/video, and apply the aforementioned techniques for the applications of computational photograph and realistic graphics rendering.
英文关键词: intrinsic images;energy optimization;texture replacement;shadow detection;