Deblur-CycleGAN: A Generative Cyclic Approach for Image Blind Motion Deblurring

In this paper, we propose an end-to-end generative adversarial network (GAN) for single image blind motion deblur-ring, which we called Deblur-CycleGAN. Inspired by the cyclic nature of the original CycleGAN, we perform single image blind motion deblurring in similar fashion while presenting motion deblurring as a cycle-consistent approach. Our proposed method achieves the best qualitative and quantitative results in comparison with existing state-of-the-art methods on GoPro dataset. We also explore the industrial aspect of motion deblurring in wind turbines (WT) with surface cracks on turbine blades. We collect 700 high-resolution images of faulty WT blades via UAV, which we called Turbine Blade dataset. Finally, we compare the performance of our proposed method against existing methods on Turbine Blade dataset and show that our proposed approach achieves the best performance both qualitatively and quantitatively.