Unsupervised image rain removal method based on attention generative adversarial network

Abstract

The invention discloses an attention generative adversarial network-based unsupervised image rain removal method, relates to the field of computer vision, and mainly relates to a neural network model capable of carrying out unsupervised learning and improving a picture rain removal effect. The problem that paired data are difficult to acquire when the generative adversarial network is trained can be effectively solved, and an attention mechanism is introduced, so that the network focuses on a rain area when processing an image, and a more ideal rain-free image is output. The method comprises the steps of 1, constructing a data set; 2, building a convolutional neural network; 3, training the network; 4, putting the process into actual use. Two discriminators of the original cyclic adversarial generative network are replaced by one discriminator, the network is simplified, the calculation amount is reduced, the discrimination capability of the discriminator is improved, and the image recovery capability of the generator is further improved. Therefore, the network focuses on a rain area when processing the image, and a more ideal rain-free image is output.

Description

technical field [0001] The present invention relates to the field of computer vision, and mainly relates to a neural network model capable of performing unsupervised learning and improving the effect of removing rain from pictures. The visibility of images processed by the network model is significantly improved. It is mainly used in image style conversion and data enhancement of automatic driving target recognition. Background technique [0002] Rain can degrade the visual quality of captured images and videos. Rain streaks (especially in heavy rain) can severely obscure the background. The accumulation of rainwater makes the distant rain streaks impossible to see alone, and together with the water particles, forms a veil on the background, which greatly reduces the contrast and visibility of the background. Human vision and many computer vision algorithms suffer from this image corruption because common computer algorithms assume clear weather and do not separately accou...

AI Technical Summary

Problems solved by technology

[0004] Aiming at the above problems, the present invention proposes an unsupervised rain removal method for the confrontation generation network that introduces the attention mechanism, which can effectively overcome the difficulty in obtaining paired data required for training the generation confrontation network, and introduces the attention mechanism, Make the network focus on the rainy area when processing the image, and output a more ideal rain-free image

Images