Single Band NIR-to-RGB Image Colorization using Attention Guided Conditional GAN

NIR images, capturing light beyond visible spectrum, are useful in remote sensing, astronomy, etc. Converting them to RGB format makes them better to interpret for humans. This paper introduces a novel approach for NIR to RGB image colorization using GAN. Our primary contribution is a Conditional Generative Adversarial Network (cGAN) with an enhanced Attention UNet serving as its generator. With this model, we have achieved an Angular Error of 15.55, PSNR of 16.41 and SSIM of 0.69 which is comparable to the other prepared models’ like pix2pix & Attention UNet as well as more visually appealing than all of them.