Image Restoration

We propose a novel deep defocus deblurring network that leverages the strength and overcomes the shortcoming of light fields. This strategy is proved to be highly effective and able to achieve the state-of-the-art performance both quantitatively and qualitatively on multiple test sets. Extensive ablation studies have been conducted to analyze the effect of each network module to the final performance.

We propose a novel convolutional neural network architecture AIFNet for removing spatially-varying defocus blur from a single defocused image. To remedy the lack of real defocused image datasets, we leverage light field synthetic aperture and refocusing techniques to generate a large set of realistic defocused and all-in-focus image pairs depicting a variety of natural scenes for network training.