Lightweight neural network single-image defogging method based on multi-scale convolution

Abstract

The invention discloses a lightweight neural network single-image defogging method based on multi-scale convolution. The method comprises the following steps: (1) dividing a data set into a training set, a verification set and a test set; (2) training the proposed network model by using the training set and the verification set; (3) testing the trained model by using the test set; and (4) evaluating the model by adopting a measurement standard. According to the method, the multi-scale defogging module is specially designed and used for extracting the picture information under each scale, whichhas an important influence on the final defogging result; model parameters are reduced as much as possible under the condition that information of all scales is fully considered, the size of the model is compressed, and then the calculation complexity of the model is reduced; a defogging effect with a higher peak signal-to-noise ratio is achieved, and meanwhile defogging of a single picture of transmissivity and global atmospheric light does not need to be estimated.

Description

technical field [0001] The invention relates to the technical field of image processing, in particular to a lightweight neural network single image defogging method based on multi-scale convolution. Background technique [0002] As a natural phenomenon, fog has not only affected people's visual experience for a long time, but also the captured fog image is not conducive to the further processing of images, which seriously affects the further development of other applications in the field of image processing, such as: image classification, object tracking, etc.; Images captured in these environments often suffer from low contrast, halos, and color shifts that limit the visibility of the images. In addition, in the field of computer vision, many computer vision and image processing algorithms are not robust to foggy images, and the existence of fog hinders further processing of images. Whether it is to improve people's visual experience or to promote the development of relate...

AI Technical Summary

Problems solved by technology

[0010] Most of the above various defogging methods have to overcome the difficulty of accurately estimating the transmittance t and the global atmospheric light A; a small number of image-to-image networks are increasing in size with the improvement of the defogging effect. Causes computational difficulties for model training and application

In addition, despite the model size and growth to a very high level, there are still details missing at some scales

Images