Multi-channel and Multi-scale Attention Mechanism Single Image Rain Removal Method
In order to remove rain streaks and rain lines that appear in images affected by rainy weather,this paper proposes a single image removal method based on a multi-channel and multi-scale attention mechanism.Through multi-scale feature extraction and network fusion,the rain streaks and rain line features of different channels in a multi-scale convolutional neural network are extracted.Firstly,bilateral filtering is used for image decomposition.Then,multi-scale feature extraction and fusion are performed on the low-frequency part,and regional attention is used to further extract the feature information of the image.At the same time,multi-scale feature extraction convolutional neural networks are used for feature learning on the high-frequency part.Finally,the two parts are added together to obtain a clearer image with more thorough removal of rain streaks and rain lines.Compared among other rain removal methods on the synthetic dataset and the real dataset,the experimental results show that the image obtained in this paper after removing rain streaks and rain lines is clearer,and some areas of the image have less loss of details,improved image quality after rain removal,thereby enhancing application effects and performance in fields such as image processing,computer vision,and machine learning.
NETL
NSTL
万方数据
