An image rain removal method based on multi-scale progressive fusion

Abstract

The invention discloses an image rain removal method based on multi-scale progressive fusion, which includes pyramid decomposition of rain images, association learning of rain stripes, and progressive fusion and reconstruction of multi-scale features. In the pyramid decomposition process of rain images, Gaussian sampling operators of different scales are used to sample and decompose the original rain images; in the association learning process of rain stripes, the global texture feature association is learned by using non-local network; in the multi-scale feature In the process of progressive fusion and reconstruction, the multi-scale pyramid network is used to process the features of corresponding scales, and the multi-scale rain streak information is gradually fused to assist the feature expression of the highest pyramid layer, so as to realize the multi-scale fusion of rain streak information and reconstruct the residual rain image, and then subtract the residual image from the rain image to obtain a rain-free image. The invention effectively utilizes the correlation of rain streaks in images of the same scale and different scales, models the rain streaks more accurately, and achieves a better effect of removing rain.

Description

technical field [0001] The invention belongs to the technical field of digital images, and relates to an image deraining method, in particular to an image deraining method based on multi-scale progressive fusion. Background technique [0002] The image or video data obtained in rainy days suffers from severe degradation, which greatly affects the quality and readability of the image or video content, thereby interfering with the accuracy of downstream high-level computer vision tasks. Therefore, single image rain streak removal is a fundamental process to improve the visual experience of images, and it is also an important preprocessing step for many computer vision tasks (such as segmentation, detection, tracking, recognition, classification, etc.). [0003] Since deep learning, especially convolutional neural network (CNN) has achieved good results in the field of image processing, image rain removal methods based on deep learning have been proposed and developed rapidly i...

AI Technical Summary

Problems solved by technology

Due to insufficient development and utilization of multi-scale rain streak information, existing deraining methods cannot produce ideal inpainting results for complex rainfall scenes

Images