Strip distortion is an important factor in video image quality reduction,but current stripe distortion removal algorithms usually require the priori knowledge of the known stripe directions,etc.How to automatically judge and procure stripe distortion images and their stripe directions then have the research significance.We propose a method in this paper to detect the stripe distortion in single direction in no-reference video image,and obtain the angle between the stripe and the positive horizontal direction.The core idea of the algorithm is based on the frequency feature of the stripe image,i.e.the spectral energy of the stripe in image will gather in spectral domain space along the vicinity of the direction orthogonal to the direction of the stripe in spatial domain,and there are peaks occurring in corresponding frequency components.For the video image with stripe distortion in single direction,the experimental results show that the proposed algorithm has good detection effect,and can determine the direction of the strip rather accurately.On the other hand,the algorithm is insensitive to random noise,but still has higher detection rate against the stripe with low contrast.%条纹失真是降低视频图像质量的一个重要因素。现有的条纹失真去除算法需要已知条纹方向等先验知识，如何自动地判断和获取条纹失真图像以及条纹的方向有重要的研究意义。提出一种检测无参考视频图像中单一方向条纹失真的方法，并获取条纹与水平正方向的夹角。算法的核心思想基于条纹图像的频率特性，即图像中的条纹的频谱能量值会在频域空间中沿着与空域中条纹方向正交的方向附近聚集，且在对应的频率分量处会出现峰值。针对单一方向的条纹失真视频图像，实验结果表明，提出的算法具有很好的检测效果，且能够较准确地确定条纹的方向。另一方面，算法对随机噪声不敏感，对低对比度条纹仍具有较高的检测率。
展开▼
