摘要
针对实时视频采用单一图像特征的拼接方法难以兼顾速度和质量的问题,提出了一种实时视频快速拼接算法。综合运用图像区域二值化(LBP)纹理特征与加速健壮特征(SURF),优化了特征点提取和匹配过程。在采用加速分块特征检测(FAST)算法提取角点的基础上,先采用改进LBP特征对备选特征点进行粗匹配,再提取SURF进行精细匹配。通过2次筛选有效降低了后续特征提取的备选点数量,并减少了匹配计算的次数和维度,从而提高了总体运算速度,可满足一定分辨率下实时视频拼接要求。
Aimed at the problem about difficult to ensure the speed and the quality simultaneously by the real time video stitching method with the single image feature, a fast real time video stitc hing algorithm is proposed. The image texture feature of the local binary patterns （LBP） and the speed up robust feature （SURF） are synthetically applied. The feature point extraction and the matching process are optimized. Based on the feature from accelerated segment test （FAST） al gorithm, the feature points to be selected are roughly matched with the improved LBP feature, and then the SURFs are accurately extracted to be matched. With two screenings, the point number to be selected in the future feature extraction is reduced, and the number and the dimen sion of the matching calculation are decreased, and the calculating speed is enhanced. Thus, the real time video stitching requirement is met under a certain resolution.
作者
卢旻昊
邱文嘉
董小舒
LU Minhao;QIU Wenjia;DONG Xiaoshu(The 28th Research Institute of China Electronics Technology Group Corporation, Nanjing 210007, China)
出处
《指挥信息系统与技术》
2018年第2期85-88,共4页
Command Information System and Technology
