The existence of shadow leads to the degradation of the image qualities and the defect of ground object information.Shadow removal is therefore an essential research topic in image processing filed.The biggest challen...The existence of shadow leads to the degradation of the image qualities and the defect of ground object information.Shadow removal is therefore an essential research topic in image processing filed.The biggest challenge of shadow removal is how to restore the content of shadow areas correctly while removing the shadow in the image.Paired regions for shadow removal approach based on multi-features is proposed, in which shadow removal is only performed on related sunlit areas.Feature distance between regions is calculated to find the optimal paired regions with considering of multi-features(texture, gradient feature, etc.) comprehensively.Images in different scenes with peak signal-to-noise ratio(PSNR) and structural similarity(SSIM) evaluation indexes are chosen for experiments.The results are shown with six existing comparison methods by visual and quantitative assessments, which verified that the proposed method shows excellent shadow removal effect, the brightness, color of the removed shadow area, and the surrounding non-shadow area can be naturally fused.展开更多
1 Introduction With the rapid progress of Artificial Intelligence(AI)technology in object detection and face recognition,deep learning methods for face mask wearing detection have become increasingly mature and contin...1 Introduction With the rapid progress of Artificial Intelligence(AI)technology in object detection and face recognition,deep learning methods for face mask wearing detection have become increasingly mature and continuously take into account the needs of efficiency and accuracy.However,these conventional detection methods mostly ignore the complexity of real-world application scenarios,such as extremely darkness and gloomy weather.These unfavorable conditions lead to a series of image degradations that seriously hamper machine vision tasks.Although camera parameter adjustment,auxiliary lighting,or pre-processing enhancement[1]can weaken these negative effects to some extent to promote the detection,they will also result in increased hardware and time costs.展开更多
基金Supported by the National Natural Science Foundation of China (No. 41971356, 41701446)the Open Fund of Key Laboratory of Urban Land Resources Monitoring and Simulation,Ministry of Natural Resources (No. KF-2022-07-001)。
文摘The existence of shadow leads to the degradation of the image qualities and the defect of ground object information.Shadow removal is therefore an essential research topic in image processing filed.The biggest challenge of shadow removal is how to restore the content of shadow areas correctly while removing the shadow in the image.Paired regions for shadow removal approach based on multi-features is proposed, in which shadow removal is only performed on related sunlit areas.Feature distance between regions is calculated to find the optimal paired regions with considering of multi-features(texture, gradient feature, etc.) comprehensively.Images in different scenes with peak signal-to-noise ratio(PSNR) and structural similarity(SSIM) evaluation indexes are chosen for experiments.The results are shown with six existing comparison methods by visual and quantitative assessments, which verified that the proposed method shows excellent shadow removal effect, the brightness, color of the removed shadow area, and the surrounding non-shadow area can be naturally fused.
基金funded by the National Natural Science Foundation of China(Grant Nos.41971356,41701446)the Open Fund of Key Laboratory of Urban Land Resources Monitoring and Simulation,Ministry of Natural Resources(KF-2022-07-001).
文摘1 Introduction With the rapid progress of Artificial Intelligence(AI)technology in object detection and face recognition,deep learning methods for face mask wearing detection have become increasingly mature and continuously take into account the needs of efficiency and accuracy.However,these conventional detection methods mostly ignore the complexity of real-world application scenarios,such as extremely darkness and gloomy weather.These unfavorable conditions lead to a series of image degradations that seriously hamper machine vision tasks.Although camera parameter adjustment,auxiliary lighting,or pre-processing enhancement[1]can weaken these negative effects to some extent to promote the detection,they will also result in increased hardware and time costs.
