Image fusion is the key step to improve the performance of object detection in polarization images. We propose an unsupervised deep network to address the polarization image fusion issue. The network learns end-to-end mapping for fused images from intensity and degree of linear polarization images, without the ground truth of fused images. Customized architecture and loss function are designed to boost performance. Experimental results show that our proposed network outperforms other state-of-the-art methods in terms of visual quality and quantitative measurement.
ASJC Scopus subject areas
- Atomic and Molecular Physics, and Optics