Reflection Removal via Realistic Training Data Generation

Abstract

We present a valid polarization-based reflection contaminated image synthesis method, which can provide adequate, diverse and authentic training dataset. Meanwhile, we enhance the neural network by introducing the reflection information as guidance and utilizing adaptive convolution kernel size to fuse multi-scale information. We demonstrate that the proposed approach achieves convincing improvements over state of the arts.

@incollection{pang2020reflection,
  title={Reflection Removal via Realistic Training Data Generation},
  author={Pang, Youxin and Yuan, Mengke and Fu, Qiang and Yan, Dong-Ming},
  booktitle={ACM SIGGRAPH 2020 Posters},
  pages={1--2},
  year={2020}
}