Densenet-Based Multi-scale Recurrent Network for Video Restoration with Gaussian Blur

摘要

Video cameras are now commonplace and available, and videos can be obtained almost everywhere at anytime. However, due to turbulence or thermal effects of air, blurring occurs during image acquisition. Removing these artifacts from the blurry recordings is a highly ill-posed problem as neither the sharp image nor the blur kernel is known. Propagating information between multiple consecutive blurry observations can help restore the desired sharp video. In this work, we propose an efficient approach to produce a significant amount of realistic training data and introduce a novel multi-scale recurrent network architecture to deblur frames taking temporal information into account. The experimental results demonstrate the effectiveness of the proposed method.