Researchers at Chinese smartphone company Xiaomi have devised an artificial intelligence (AI) system that corrects details and colors in poorly exposed photos by segmenting the image into multiple "sub-images," each associated with a local exposure, that it consequently uses to retouch the original input photo.
According to the team, "Inspired by luminosity masks usually applied by professional photographers...we develop[ed] a novel algorithm for learning local exposures with deep reinforcement adversarial learning" via a generative adversarial network.
Following image segmentation, the DeepExposure AI concatenates and processes the input low-resolution, sub-images, and direct fusion through a policy network that computes each's local and global exposures.
Afterwards, a value function assesses overall quality and the sub-images are combined with the input image.
DeepExposure was trained in Google's TensorFlow framework, and it has outperformed state-of-the-art algorithms in key metrics, consistently restoring most details and styles in original images while enhancing brightness and colors.
View Full Article
Abstracts Copyright © 2018 Information Inc., Bethesda, Maryland, USA
No entries found