Wavelet-based direction perception attention image rain removal method, system and equipment

Abstract

The invention discloses a wavelet-based direction perception attention image rain removal method, system and device. The method comprises the steps of wavelet decomposition and embedded expression of a rain image, vector decomposition and feature expression based on direction perception, reverse wavelet background reconstruction, and gradient weight joint optimization of an image space and a color channel. In the wavelet decomposition and embedding expression process of the rainwater image, Haar wavelet decomposition is utilized to obtain wavelet components of four different frequencies of low frequency, horizontal frequency, vertical frequency and diagonal frequency. In a vector decomposition and feature expression process based on direction perception, horizontal and vertical components are separated and reconstructed from low-frequency and diagonal rainwater components through an attention learning strategy of direction perception. In the reverse wavelet background reconstruction process, rainwater interference of different components is removed in different stages of the network, wavelet components corresponding to a clean background image are obtained, and the background image is obtained through corresponding Haar wavelet inverse transformation. According to the invention, background texture restoration with higher quality is realized.

Description

technical field [0001] The invention belongs to the technical field of digital images, relates to a method, system and device for removing rain from images, and in particular relates to a method, system and device for removing rain from images based on wavelet direction perception attention. Background technique [0002] Under the condition of low visibility in rainy days, the image quality will be disturbed to varying degrees, including image content occlusion, brightness change, contrast distortion, etc., which greatly affects the visibility of image information, thereby restricting the processing accuracy of downstream high-level vision tasks. Therefore, image deraining is an important preprocessing step for many computer vision tasks such as segmentation, detection, tracking, recognition, classification, etc. [0003] In recent years, image rain removal methods based on deep learning have been proposed one after another. In 2017, Fu et al. first built a three-layer CNN ...

AI Technical Summary

Problems solved by technology

Although this method can remove most of the rain streaks, it does not take into account the specific direction of the rain streaks, which will damage the background texture in the vertical direction or cause color distortion.

Images