Traditional cameras can't always be relied on to offer a clear view of a subject, with adverse conditions such as rain, fog, or a lack of light reducing visibility. However, researchers from NEC and the Tokyo Institute of Technology have now used artificial intelligence to automatically combine visible and non-visible images, dramatically increasing clarity in the resulting shots.
Non-visible images, such as those produced by things like thermal or X-ray cameras, have been used in combination with traditional photography for some time, for purposes such as monitoring situations in low light or severe weather conditions. However, this has typically required manual processing and combining, and can easily result in aspects of the non-visible images getting overlooked.
Now researchers have developed a way of using AI to automatically combine these visible and non-visible images, resulting in images with greater visibility and clarity. The new technology works by examining each image to assess the degree of visibility and then automatically extracting the best areas from each image, taking into account environmental characteristics and obstacles.
The AI also detects abnormalities in the non-visible camera images and can generate multimodal images fusing the visible and non-visible detail. The resulting images are said to offer exceptionally high visibility while the regulation of enhancements avoids problems like clipped highlights and crushed blacks.
Potential uses of the new technology are said to include monitoring systems to assist with nighttime observations, and infrastructure inspection devices which could be used to detect abnormalities and issues such as cracks buildings and structures.
NEC and the Tokyo Institute of Technology will present the findings at the Symposium on Sensing via Image Information in Yokohama City, Japan, on June 7.