Bayesian statistical models based on global constraints in computer vision and image processing

我的研究计划研究新的无监督概率或基于能量的模型的使用，包括作为新奇的特定几何、功能、生理或空间自适应全局先验约束，用于恢复、重建和理解静止和运动图像。通过这种方式，本研究计划将尝试提出一个统计框架来集成多个(低或高级别)图像线索(例如，边缘、分割图、速度场、深度图等)，以提高要恢复的解决方案的准确性，或用于融合来自多种和互补的医学成像模式(SPECT与MRI、血管造影术与X光等)的多个图像线索。用于解剖对象的有效恢复或识别或3D重建，或用于对某些医学现象的整体理解。更广泛地说，本研究计划旨在设计下一代自动识别重建和理解系统

My research program investigates the use of new unsupervised probabilistic or energy-based models including, as novelty, specific geometrical, functional, physiological or spatial adaptive global a priori constraints for restoring, reconstructing and understanding still and moving images.  In this way, this research program will attempt to propose a statistical framework to integrate multiple (low or higher level) image cues (e.g. edges, segmentation map, velocity field, depth map, etc.)  to improve the accuracy of the solution to be recovered or  to fuse multiple image cues coming from  multiple and complementary medical imaging modalities (SPECT with MRI, angiographic with Xrays, etc.) for the efficient restoration or the recognition or 3D reconstruction of anatomical objects or for the overall understanding of certain medical phenomenons. More generally, this research program aims at designing automatic recognition reconstruction and understanding systems of the next generations