AI Developed by Russian Scientists Will Colorize Nighttime Satellite Images
Researchers in Nizhny Novgorod have patented a new method for restoring and colorizing black-and-white images of large cities.
Scientists at Nizhny Novgorod State University have received a patent for a new technology, the university’s press service said. The researchers have developed a method for colorizing nighttime images of cities and urban agglomerations. The technology is intended for economic and logistics research, more accurate urban planning, and assessing environmental impact.
One Neural Network Per Color
The neural network processes black-and-white images captured by space satellites in several stages. First, each pixel is mapped to its real-world location, and the level of urban development in that area is calculated. The image is then colorized using the three RGB channels. Red represents brightness, green shows the development density coefficient, and blue indicates the degree of deviation from it.
Next, the image is divided into squares, with each color channel processed by a separate neural network. All three outputs are then combined into a single multispectral image. The development addresses a pressing challenge in the analysis of land use within urban environments.
