基于顶点轮廓的鲁棒性抗前景干扰稳像算法  

Robust Image Stabilization Method Against Foreground Interference Based on Vertex Contour

作　　者：尹丽华 康亮[1] 刘士建[2] YIN Li-hua;KANG Liang;LIU Shi-jian(Shanghai Polytechnic University,Shanghai 201209,China;CAS Key Laboratory of Infrared System Detection and Imaging Technology,Shanghai Institute of Technical Physics of the Chinese Academy of Sciences,Shanghai 200083,China)

机构地区：[1]上海第二工业大学,上海201209 [2]中国科学院上海技术物理研究所中国科学院红外探测与成像技术重点实验室,上海200083

出　　处：《计算机仿真》2023年第9期184-190,196,共8页Computer Simulation

基　　金：教育部科技发展中心产学研创新基金项目(2018C01059);“十三五”国防预研项目(Jzx2016-0404/Y72-2)。

摘　　要：为了解决复杂运动前景对稳像精度干扰的难题,提出了一种基于顶点轮廓的鲁棒性抗前景干扰稳像算法。先对输入视频检测FAST特征,将图像划分为m*n的网格,结合基于金字塔的Lucas-Kanade光流算法得到特征匹配对;并根据运动矢量的相似性,利用中值滤波器f1生成基于网格顶点的稀疏运动矢量。通过时间分析法识别出具有不连续运动矢量的网格顶点,剔除“离群”顶点处的运动矢量,并利用中值滤波器f2对缺失的运动矢量进行补全,实现运动矢量的空间平滑。最后形成顶点轮廓,并结合多路径平滑策略实现图像稳定。实验结果表明:当处理大范围和多运动前景干扰的情况时,上述算法比传统稳像算法的PSNR值分别高14.9%和13.2%,视频的过渡更加平滑,且具有更好的稳像精度和鲁棒性,运行效率也更高。In order to solve the problem of interference of complicated motion foreground to video stabilization accuracy,a robust image stabilization method against foreground interference based on vertex contour is proposed.Firstly,FAST features are detected for the input video.Then,the image is divided into m*n grids,and the feature matching pairs are obtained by combining the pyramidal Lucas-Kanade optical flow algorithm.According to the similarity of motion vectors,the median filter fl is used to generate sparse motion vectors based on mesh vertices.Then,the mesh vertices with discontinuous motion vectors are identified by time analysis method,the motion vectors at outlier vertices are removed,and the median filter f2 is used to complete the missing motion vectors to achieve spatial smoothing of motion vectors.Finally,the vertex contour is formed and the multipath smoothing strategy is used to stabilize the image.Experimental results show that when dealing with large range and multi-motion foreground interference,the PSNR value of the proposed algorithm is 14.9%and 13.2%higher than that of the traditional image stabilization algorithm,and the video transition is smoother,with better image stabilization accuracy and robustness,and higher operation efficiency.

关 键 词：稳像 顶点轮廓 抗前景 鲁棒 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]