Perceptually and probabilistically guided multimedia

在接下来的五年里，我将开发生物和概率动机的多媒体系统，包括以下新的进步： (i)3D医学图像分析中的概率和感知因素，扩展技术，如随机游走和高斯形状建模，以更好地表示与婴儿大脑、脊髓等相关的特定医疗状况，从而改进分段技术。 (ii)在3D点云简化过程中融入轮廓的感知。 (iii)扩展为网格数据结构开发的无线3D传输策略，以使用点云表示。 (iv)使用简单、广泛可用的对象（如地球仪）校准和同步多相机环境，并结合传感器数据创建动态集成3D场景。 (v)设计优化的周边视觉立体显示器。 (vi)感知引导的运动捕捉（MoCap）数据压缩，目标是即使在手持蜂窝设备上也能实现实时解压缩。 我工作的意义在于： (a)强大的3D医学图像分割和地面实况生成算法。 (b)在3D多媒体传输和可视化中降低带宽、内存需求并优化感知质量;应用于全景3D电视。 (c)在数据包丢失的情况下，采用稳健的3D传输策略;通过无线设备改进3D教育娱乐。 (d)强大的摄像机网络校准和智能协调，实现更好的3D自由视点视频捕捉。 (e)在寓教于乐中开发更好的交互式3D环境。

Over the next five years I will develop biologically and probabilistically motivated multimedia systems that include the following novel advancements: (i) Probabilistic and perceptual factors in 3D medical image analysis, extending techniques, such as random walks and Gaussian shape modeling, to better represent specific medical conditions related to infant brain, spinal cord etc., and thereby improve segmentation techniques. (ii) Incorporate the perception of silhouettes in the process of 3D point cloud simplification. (iii) Extend wireless 3D transmission strategies developed for mesh data structures to work with point-cloud representations. (iv) Calibration and synchronization of multi-camera environments using simple, widely available objects (such as a globe) and combining sensor data to create a dynamic integrated 3D scene. (v) Designing optimized peripheral vision stereo displays. (vi) Perceptually guided Motion Capture (MoCap) data compression, with the goal of making real-time decompression feasible even on hand-held cellular devices. The significance of my work will lie in: (a) Robust 3D medical image segmentation and ground truth generation algorithms. (b) Reducing bandwidth, memory requirements and optimizing perceptual quality in 3D multimedia transmission and visualization; with applications in panoramic 3D TVs. (c) Robust 3D transmission strategies under packet loss; resulting in improved 3D edutainment over wireless devices. (d) Robust calibration and intelligent coordination of a network of cameras, allowing better 3D free-viewpoint video capture. (e) Developing better interactive 3D environments in Edutainment.