Learn to model blurry motion via directional similarity and filtering

作者：

Highlights：

• A hybrid framework to extract optical flow from blurry footages.

• A CNN component for image deblurring.

• A learnable directional filtering layer encodes the angle and distance similarity between blur and image properties.

• Two synthetic ground truth sequences for the blurry scenes.

摘要

•A hybrid framework to extract optical flow from blurry footages.•A CNN component for image deblurring.•A learnable directional filtering layer encodes the angle and distance similarity between blur and image properties.•Two synthetic ground truth sequences for the blurry scenes.

论文关键词：Optical flow,Convolutional Neural Network (CNN),Video/image deblurring,Directional filtering

论文评审过程：Received 21 July 2016, Revised 7 February 2017, Accepted 17 April 2017, Available online 22 April 2017, Version of Record 21 November 2017.

论文官网地址：https://doi.org/10.1016/j.patcog.2017.04.020