TY - JOUR T1 - Total Variation Regularization Low-Rank Decomposition Based Tensor Model for Video Rain Streaks Removal AU - Lu , Xinghan AU - Zheng , Yuhui AU - Zhang , Jianwei JO - Journal of Information and Computing Science VL - 1 SP - 052 EP - 063 PY - 2024 DA - 2024/01 SN - 16 DO - http://doi.org/ UR - https://global-sci.org/intro/article_detail/jics/22379.html KW - Rain removal, tensor model, total variation, low rank, ADMM. AB -

With superior real-time and storage performance, outdoor computer vision systems have high application value in traffic, public security, identification detection and other fields, but the captured images are affected by environmental factors such as outdoor rainfall, which have obscuration or missing problems and are not conducive to the processing and application of post-level systems. To this end, this paper proposes a tensor model based on total variation regularization low-rank decomposition for video rain streaks removal. Considering the influence of moving objects in the video image on the low-rank structure of the video background, the rainy video is decomposed into static background, dynamic objects and rain streaks, and their a priori characteristics are analyzed separately, combined with the corresponding low-rank characteristics or sparse characteristics to construct a tensor model, and the targets are extracted through low-rank decomposition, and then the rain removal is completed. The proposed tensor model is solved by the alternating direction multiplier method (ADMM), and extensive experiments are carried out on synthetic and real data sets. The results show that the proposed method can effectively remove rain streaks from video images while retaining more background details under dynamic background conditions. Compared with related advanced methods, the proposed method has advantages in three comprehensive quantifiers, namely, peak signal-to-noise ratio, structural similarity and residuals.