Cross-Camera Deep Colorization

In this paper, we consider the color-plus-mono dual-camera system and propose an end-to-end convolutional neural network to align fuse images from it in efficient cost-effective way. Our method takes cross-domain cross-scale as input, consequently synthesizes HR colorization results facilitate trade-off between spatial-temporal resolution color depth single-camera imaging system. contrast previous methods, ours can adapt monochrome cameras with distinctive resolutions, rendering flexibility robustness practical applications. The key ingredient of our is a cross-camera alignment module that generates multi-scale correspondences for image alignment. Through extensive experiments on various datasets multiple settings, validate effectiveness approach. Remarkably, consistently achieves substantial improvements, i.e., around 10dB PSNR gain, upon state-of-the-art methods. Code at: https://github.com/THU-luvision .