监控视频的高动态范围化及其在辐射域处理

Video surveillance is vital in the fields of forensics and criminal investigation. The usefulness of such a system for these tasks is strictly determined by video quality which is affected by camera performance, lighting condition, motion blurring, distance, as well as dynamic range. To cope with the problem that most video surveillance systems cannot distinguish features with sufficient clarity, a novel method is presented to generate high dynamic range video without using troublesome multi-exposures and changing current video system. A method for determining camera response function from varying illuminated images will be studied. Further, it is proposed to process images in radiance domain. It is expected to achieve better results than the conventional intensity-domain image processing because this idea follows the imaging pipeline and avoid saturation yielded by wide dynamic range of scenes. The key target of the project is to provide clear video sequence based on cost-effective video system in order to see everything within the field of view at any time, and develop an excellent image/video processing tool for video enhancement in law enforcement, transportation authority and other departments using video surveillance. The research conducted in this project will contribute to the fields of image/video processing, computer graphics and computer vision.

视频监控已成为公安部门取证和破案的重要技术手段。针对目前视频监控系统所拍摄的图像分辨率低，加上光照变化、运动模糊和距离的不确定，导致监控画面难以辨别的问题，本项目提出无需多次曝光，无需对现有监控视频系统做任何改动的情况下，实现高动态范围视频的研究；探讨不同光照下摄像头响应函数计算方法；提出在辐射域进行图像处理，如去噪、去模糊及运动图像合成。与传统图像复原模型相比，本方法符合成像原理，避免了因为传统模型不精确以及场景动态范围过宽造成图像像素饱和所带来的一系列处理问题。通过本项目的研究，以期把人眼看不清的图像清晰化，为开发一套具有自主版权的图像/视频处理系统打下坚实的基础，为公安系统，交通管理等需要视频监控的部门带来便利。本项目研究成果对图像/视频处理，计算机图形学及计算机视觉等领域有较高的学术价值。

视频监控已成为公安部门取证和破案的重要技术手段。针对目前视频监控系统所拍摄图像的动态范围低，导致监控画面难以辨别的问题，本项目提出无需多次曝光，无需对现有监控视频系统做任何改动的情况下，实现高动态范围视频的研究；在本项目中，我们进行了以下研究：1）针对同一场景不同光照的图像相关性和不变性展开研究；2）探讨了不同光照下摄像头响应函数估计方法；3）具有边缘保持性的图像滤波研究；4）低光照视频增强研究；5）具有结构保持性的自适应去噪方法研究；6）利用多帧信息进行移动目标检测及跟踪；7）图像配准及融合中的去鬼影研究；8）细节保留的色调映射研究；9）通过改变辐射值进行信息融合以生成高动态范围视频。基于本项目的相关理论研究成果，开发了一套具有自主版权的高动态范围视频监控演示系统，并可用于公共安全、交通管理等需要视频监控的部门。本项目研究成果对图像/视频处理，计算机图形学及计算机视觉等领域有较高的学术价值。