M-CBN: Manifold constrained joint image dehazing and super-resolution based on chord boosting network

Pengyu Wang, Hongqing Zhu, Han Zhang, Nan Wang

Abstract: Highlights•Joint image dehazing and super-resolution learning promotes high-resolution image detail restoration especially for excessively degraded hazy images.•Cross-linked back-projection between high- and low-resolution features and cross-resolution spatial attention can recover missing spatial information.•Detail information can be boosted by dense high- and low-frequency feature cross-collaboration.•The static and dynamic discriminators jointly optimize this dehazing and super-resolution task.