基于Kinect视频的图像增晰与检测算法研究

Research on Image Enhancement and Detection Algorithm Based on Kinect Video

图像中的元素目标检测和图像增晰是计算机在图像计算、视觉表达中的重要任务,尤其是在视频监控、导航、智能交通、医药等领域内有着重要的应用价值。利用HOG/HOD框架和SSD目标检测算法,以及中值滤波、双边滤波、形态学等图像增晰方法,对不同的数据集进行实验与测试,旨在通过Kinect深度视频在目标检测和图像增晰方面有所增进。研究发现:HOG算法与SSD算法均能理想地完成目标检测和图像增晰,HOG算法耗时性能优越,SSD算法在检测精确率和召回率上较HOG算法更优越;利用中值滤波对RGB图像进行增晰,能够有效改善图像的色彩像素噪点等瑕疵。

Element target detection and image enhancement are important in computer calculation and visual expression.Especially,they have important application values in fields such as video surveillance,robot navigation,intelligent transportation,and medicine.Using HOG/HOD framework and SSD target detection algorithm,as well as image enhancement algorithms such as median filtering,bilateral filtering,morphology,experiments and tests are carried out on different data sets,aiming at improving target detection and image enhancement through Kinect in-depth video.The results show that both the HOG algorithm and the SSD algorithm can ideally complete the target detection and image enhancement.The HOG algorithm is less time-consuming than the SSD algorithm,but the SSD algorithm is better than the HOG algorithm in detection accuracy and recall;using the median filter to perform the RGB image enhancement can effectively improve the color,pixel,noise and other defects of the image.