It is difficult to recover the motion field from a real-world footage given a mixture of camera shake and other photometric effects. In this paper we propose a hybrid framework by interleaving a Convolutional Neural Network (CNN) and a traditional optical flow energy. We first conduct a CNN architecture using a novel learnable directional filtering layer. Such layer encodes the angle and distance similarity matrix between blur and camera motion, which is able to enhance the blur features of the camera-shake footages. The proposed CNNs are then integrated into an iterative optical flow framework, which enable the capability of modeling and solving both the blind deconvolution and the optical flow estimation problems simultaneously. Our framework is trained end-to-end on a synthetic dataset and yields competitive precision and performance against the state-of-the-art approaches.
Learn to Model Blurry Motion via Directional Similarity and Filtering
2018-03-01
Li , W , Chen , D , Zhihan , L , Yan , Y & Cosker , D 2018 , ' Learn to Model Blurry Motion via Directional Similarity and Filtering ' , Pattern Recognition , vol. 75 , pp. 327-338 . https://doi.org/10.1016/j.patcog.2017.04.020
Aufsatz (Zeitschrift)
Elektronische Ressource
Englisch
Convolutional Neural Network (CNN) , Directional filtering , Optical flow , Video/image deblurring , /dk/atira/pure/subjectarea/asjc/1700/1712 , Software , /dk/atira/pure/subjectarea/asjc/1700/1711 , Signal Processing , /dk/atira/pure/subjectarea/asjc/1700/1707 , Computer Vision and Pattern Recognition , /dk/atira/pure/subjectarea/asjc/1700/1702 , Artificial Intelligence
DDC: | 629 |
NTIS | 2019
|NTRS | 2019
|Robustness of noisy and blurry images segmentation
British Library Online Contents | 2009
|On high-speed rail´s beaten and blurry path
IuD Bahn | 2011
|