摘要
针对Kinect传感器获取的深度图像中存在大量噪声以及深度信息缺失导致的空洞问题,提出一种基于时空域数据融合的深度图像修复算法。首先,对配准后的深度图像利用卡尔曼滤波使跳变深度值趋于平稳,并采用阈值分割法得到待修复区域;其次,计算待修复边界所有像素点的时空域置信度,对时空域置信度最大的像素点计算其时域和空域深度数据,并根据时空域置信度为时空数据分配权值进行数据融合,实现像素点的修复;最后,待修复边界改变,迭代执行上一步直至图像修复完成。实验结果表明:与传统修复算法相比,基于时空域数据融合的Kinect深度图像修复算法的深度图峰值信噪比更高、均方根误差更小,图像质量更好。
Aiming at a great number of problems which caused by a large number of noises in the depth image acquired by Kinect sensor and the lack of depth information,a depth image restoration algorithm which is based on spatio-temporal data fusion is proposed.Firstly,it deals with the depth image of registration through the Kalman filter to stabilize the hop depth value,and obtain the region to be repaired by the threshold segmentation method.Secondly,the space-time domain reliability of all pixels in the boundary to be repaired is calculated.The time domain and spatial depth date of the pixel with the highest reliability is calculated,and the data fusion which is accomplished in accordance with the space-time domain reliability for the spatio-temporal data distribution weight has achieved the pixel point repair.Finally,after the boundary to be repaired to be changed,an interation method to perform the previous step until the image restoration is successful.The experimental results show that compared with the traditional repair algorithm,the depth image restoration algorithm makes a peak signal-to-noise ratio more accurate,root means square error less and image quality better.
作者
林玲
陈姚节
郭同欢
LIN Ling;CHEN Yao-jie;GUO Tong-huan(College of Computer Science and Technology,Wuhan University of Science and Technology,Wuhan 430065,China;Hubei Province Key Laboratory of Intelligent Information Processing and Real-time Industrial System,Wuhan University of Science and Technology,Wuhan 430065,China;Metallurgical Industry Process National Virtual Simulation Experimental Teaching Center,Wuhan University of Science and Technology,Wuhan 430065,China)
出处
《科学技术与工程》
北大核心
2019年第30期215-220,共6页
Science Technology and Engineering
关键词
时空域置信度
时空域数据
数据融合
深度图像修复
space-time domain reliability
space-time domain data
data fusion
pictural depth restoration
