Single image defogging algorithm based on deep learning

Abstract

The invention relates to a single image defogging algorithm based on deep learning, comprising the following steps: a large number of bright and fogless images are obtained as a fogless image set in a training sample, and fog interference of different concentration is applied to the fogless image set through simulation software to generate a foggy image set; the fogless image set and the foggy image set are converted to an HDF5 format to get a training sample and a test sample; the training sample and the test sample are input to a deep convolution network of which the parameters are set, the deep convolution network is trained until the cost loss reaches a certain extent and the number of iterations reaches a maximum, and a trained model is obtained; and finally, foggy images are input to the trained model, and fogless images are restored directly. The invention provides an end-to-end convolution neural network which can restore fogless images directly from foggy images, and estimation of the intermediate parameters is omitted. Moreover, color distortion of the flat area in a foggy image is avoided, natural and synthetic foggy images can be processed effectively, and a better effect can be achieved.

Description

technical field [0001] The invention relates to a convolutional neural network and image defogging technology, in particular to an end-to-end single image defogging method based on a convolutional neural network. Background technique [0002] Fog and haze are common phenomena on land and sea. In foggy and hazy weather there are many atmospheric particles of significant size. They not only absorb and scatter the reflected light of the scene, but also scatter some atmospheric light to the camera. Consequently, images acquired by the camera are degraded and generally have low contrast and poor visibility. This will seriously affect the vision system, especially the visible light vision system. Objects and obstacles in the image are difficult to detect due to the degradation of the image. This is bad for automatic video processing, such as feature extraction, object tracking and object recognition. It is also one of the leading causes of accidents in the air, at sea and on ...

AI Technical Summary

Problems solved by technology

The purpose of the image restoration algorithm is to obtain a natural and clear image with good visibility while maintaining good color restoration performance, but the processing range is limited

Images