中国计算机学会青年计算机科技论坛

CCF Young Computer Scientists & Engineers Forum

CCF YOCSEF

2012年09月11日(星期二)10:00-12:00

国防科技大学 计算机学院后楼学术厅

敬请光临

研讨会主题

YOCSEF长沙论坛2012“可视计算”研讨会

会议简介：本次研讨会拟针对可视计算领域中的数字几何处理、造型等关键问题进行研讨，阐述最新研究成果。

程  序  手  册

09:50  签到

10:00  研讨会

10:00-10:45 Session 1（主持人：程志全）

10:00-10.45

报告人：Ralph R. Martin（英国卡迪夫大学）

报告题目：Making a Bas-Relief of a Human Face from a Single Photograph

摘要：We present a system which can make a bas-relief of a human face from a single frontal photograph. In a learning step, a reference human head model is compressed to a relief algorithmically using a known method. The 3D head and the relief are then used to generate images, and a mapping of intensities is learnt from head images to relief images. When presented with a new photograph, we use the above mapping to generate a corresponding relief image. Shape-from-shading is then used to generate the corresponding bas relief. Several strategies are used to ensure high quality results. Non-photorealistic rendering is applied to the input head images and photographs to emphasise salient features and suppresses unwanted small details. We use image relighting techniques to relight the input photograph from a pair of standard directions, which helps to both overcome problems of small shadows; two reliefs are generated and averaged which further reduces undesirable artifacts, and helps to ensure the relief has a stable appearance with respect to changes in lighting and viewing direction. Our method is robust to changes of facial appearance and expression, head pose, and lighting direction.

茶歇（5分钟）

10:50-12:00 Session 2（主持人：李实英）

10:50-11:25

报告人：姜巍（国防科大）

报告题目：Skeleton-based global intrinsic symmetry detection on point cloud

摘要：We study on intrinsic symmetry detection on incomplete point cloud where large portions of the data may be missing. Taking a point cloud as input, our method first extract a curve skeleton using a technique that couples graph contraction and point clustering. The extracted skeleton is a reduced representation of the point cloud and captures the correlation between skeleton nodes and surface sample points. Intrinsic symmetry detection is then reduced to finding an intrinsic distance preserving self-mapping on the skeleton with the help of local signatures computed for the point cloud. The symmetry correspondence is transferred on the surface using the skeleton-surface correlation. The output is a serial of pair-wise symmetric regions on the input point cloud.

11:25-11:50

报告人：沈恩亚（国防科大）

报告题目：Model-Based Volumetric Data Segmentation, Labeling, and Rendering

摘要：We present an intuitive multi-component volumetric data segmenting, labeling and rendering system, whose output is driven by a given annotated semantic model. The input volume data may vary in detail from the model, and it is not annotated with semantic information, so making use of the semantic data still remains a challenging task. Our system comprises three main steps. The first is a model-inspired semi-automatic volume segmentation algorithm, which clusters components with similar or identical scalar values using Gaussian mixture models, and the initial classification would be handled by a seed region growing method to produce fine-grained part segmentations. The second, core, step progressively labels the segmented components of the volume data driven by corresponding parts in the model. Thirdly, the volume data is rendered by suitable transfer functions, using color information provided by the model. This provides a flexible and natural data exploration and visualization approach, allowing real-time interactive feedback to the user. Our experiments on various data show that our system often produces reasonable segmentations, annotations and rendering results similar to those produced by domain experts.

执行主席：程 志全  博士，国防科技大学            

执行主席：李 实英  博士，湖南大学

                    YOCSEF长沙学术委员

参加人员：计算机图形、图像和可视化方向感兴趣者

报名方式：无需注册，自由参加

紧急联系方式：Email：cheng.zhiquan@gmail.com;

Tel: 程志全（13187015816）刘晶（15116343687）
=====================================================================

讲者 Ralph R.Martin

现为，国防科大、清华大学客座教授。1982获得剑桥大学博士学位。其研究研究领域为几何计算、计算机图形学和视频处理。目前担任Computer Aided Design等4个国际刊物的编委。Ralph R Martin教授是反向工程的权威，他和匈牙利的T Varady教授，被公认为是开创该研究领域的先驱。
 
讲者 姜巍

博士生。2004、2009年在国防科技大学计算机学院分别获得工学学士、硕士学位。研究方向包括计算机图形学和数字几何处理。

讲者 沈恩亚

博士生。2010年在国防科技大学计算机学院分别获得硕士学位。研究方向包括计算机图形学和可视化。
_______________________________________________________________________________

执行主席 程志全

博士。2000年、2003年和2008年于国防科技大学计算机学院分别获得工学学士、硕士和博士学位,2010年在英国卡迪夫大学进行学术访问。主要研究方向为计算机图形学和可视计算。

执行主席 李实英

博士，2007年于日本奈良先端科学技术大学院大学获得工学博士学位。主要研究方向为计算机视觉和图形学，视频图像处理。

会场地点：国防科大计算机学院后楼学术厅。

建议路线如下：乘车到达三一大道国防科大北门、在北门的西侧接待室进行换证工作（备好身份证，入校原因为：参加计算机学院的学术会议）、进入北校门后，步行300米后向西到达计算机学院（楼顶有“银河”招牌），在学院门口，您会到达路线指示牌，到达学术厅。