Atmospheric turbulence removal using convolutional neural network
This work addresses the problem of image degradation due to atmospheric turbulence for applications like surveillance or astronomy, representing an incremental improvement over prior methods.
The paper tackles atmospheric turbulence removal in video sequences by developing a convolutional neural network based on residual learning, which deblurs, removes ripple effects, and enhances contrast, outperforming existing methods by up to 3.8% in image quality and achieving up to 23 times faster speed with GPU implementation.
This paper describes a novel deep learning-based method for mitigating the effects of atmospheric distortion. We have built an end-to-end supervised convolutional neural network (CNN) to reconstruct turbulence-corrupted video sequence. Our framework has been developed on the residual learning concept, where the spatio-temporal distortions are learnt and predicted. Our experiments demonstrate that the proposed method can deblur, remove ripple effect and enhance contrast of the video sequences simultaneously. Our model was trained and tested with both simulated and real distortions. Experimental results of the real distortions show that our method outperforms the existing ones by up to 3.8% in term of the quality of restored images, and it achieves faster speed than the state-of-the-art methods by up to 23 times with GPU implementation.