Post date: Sep 25, 2019 3:14:59 AM
https://link.springer.com/article/10.1007/s11042-018-6536-x
- Extracting multi-scale haze-relevant representations (features) from a hazy image in an unsupervised learning pipeline using a stacked autoencoder.
- Estimating a transmission map from the haze-relavant features using another deep CNN.