Extremely sparse stripe noise removal from nonremote-sensing images by straight line detection and neighborhood grayscale weighted replacement

Traditional methods of stripe noise removal based on space domain or transformation domain generally cannot handle the case where the noise is extremely sparse. To solve this problem, we propose a novel approach to accurately detect and remove the stripe noise by analyzing the directional and structural information of the stripe noise. First, we build a preselected stripe noise lines set by using local progressive probabilistic Hough transform. Subsequently, the real stripe noise lines are screened out from this set according to the feature of grayscale discontinuities. Finally, our approach uses the strategy of neighborhood grayscale weighted replacement and a local Gaussian filter to perform image destriping. Extensive experiments demonstrate that our approach proposed in this paper outperforms other recent promising methods in terms of quantitative assessments, qualitative assessments, and computing time.