Realistic Blur Synthesis for Learning Image Deblurring

Training learning-based deblurring methods demands a tremendous amount of blurred and sharp image pairs. Unfortunately, existing synthetic datasets are not realistic enough, and deblurring models trained on them cannot handle real blurred images effectively. While real datasets have recently been proposed, they provide limited diversity of scenes and camera settings, and capturing real datasets for diverse settings is still challenging. This paper analyzes various factors that introduce differences between real and synthetic blurred images, and presents a novel blur synthesis pipeline to synthesize more realistic blur. We also present RSBlur, a novel dataset with real blurred images and the corresponding sharp image sequences to enable detailed analysis on the differences between real and synthetic blur. With our blur synthesis pipeline and RSBlur dataset, we reveal the effects of different factors in blur synthesis. We also show that our synthesis method can improve the deblurring performance on real blurred images.