AIST: Improved colorization of infrared images: Color reproduct by deep learning

AIST: Improved colorization of infrared images: Color reproduct by deep learning

AIST: Infrared image

From infrared night vision images, the “color reproducibility of visible color images” has been greatly improved through deep learning.

Points of development:

Real-time infrared video colorization possible
Applied to security cameras with high visibility
Expectations for application to ecological records of nocturnal animals

Development overview:

“AIST” / Analysis and Measurement Standard Research Division / Nano Spectroscopy Measurement Research Group

We have developed a technology to colorize infrared images with visible light by “deep learning using multilayer / artificial neural / network”.

Until now, “infrared night-vision images that could only be displayed in monochrome and approximate colors”

“Displaying in colors that are very close to colors under visible light” has become possible.

https://www.aist.go.jp/aist_j/press_release/pr2020/pr20200206/pr20200206.html