摘要
针对视频监控图像中存在各类条纹噪声的问题,根据条纹特性和受干扰图像帧的频域特征,提出了一种快速检测监控录像周期性条纹的算法。根据相对距离将频谱图分成两个子块,再运用行列累积函数或阈值检测各子带是否存在异常亮点,进而确定图像帧是否存在条纹噪声。利用频率谱中异常点的对称特性可减少遍历次数,有效提高了算法的运行效率。实验结果表明,该算法对监控视频序列中的多种周期性条纹具有良好的检测效果,并提高了计算速度。
To solve the striping noise problem that occurred in the video surveillance image, this paper presents a fast method for detecting periodical stripe noise which based on stripe characteristics analyses of an image frame. Firstly, frequency spectrum is divided into two sub-bands by using relative distance. Then, detect if there is any exceptional point in each sub-band by using cumulative distribution or threshold methods. Lastly, presence or absence of stripe noise can be determined. Since the counts of scanning computation can be decreased according to symmetry property of abnormal point in the frequency spectrum, algorithm efficiency can be improved clearly. Experimental results show that the proposed algorithm can effectively detect many kinds of periodical stripe noises in the video monitoring sequence with better effectiveness and computation speed.
出处
《微型机与应用》
2011年第17期36-39,共4页
Microcomputer & Its Applications
基金
上海海事大学基金项目(No.20110043)
关键词
视频监控图像
条纹噪声
傅里叶变换
累积分布函数
video surveillance image
stripe noise
Fourier transform
the cumulative distribution function
