News and events - Tampere University of Technology

TUT Signal Processing Research Results - Better Digital Cameras

Do you look like a smurf or raw beef in the picture you just took? No worries, researchers of Laboratory of Signal Processing provide a novel state-of-the-art solution for you - "Recurrent Color Constancy".
So the colors didn't turn up that well after all...
So the colors didn't turn up that well after all...

During the FiDiPro project two project partners Tampere University of Technology (TUT) and Intel Finland developed novel technology for the problem of color constancy which has been around for decades and despite of huge amount of research and many important technologies and patents the colors are just wrong sometimes!

The novel solution was found by thinking out-of-the-box - how about if a camera has memory and it can adjust the color settings based on its viewfinder history and what was seen just before the final shot - simple but effective! The solution is based on the recent deep convolutional neural network and long-short-term memory technologies that have revolutionized artificial intelligence research during the past few years.

The new technology will be presented in the highly prestigious top-tier computer vision conference this year in Venice: International Conference on Computer Vision (ICCV) 2017:

  • Recurrent Color Constancy (Y. Qian, K. Chen, J.-K. Kämäräinen, J. Nikkanen, J. Matas), In Int. Conf. on Computer Vision (ICCV2017), 2017, Venice, Italy.
News submitted by: Joni Kämäräinen
Keywords: education and studies, science and research, services and collaboration, about tut, image and communications, signal processing, computer vision, machine learning, artificial intelligence, color constancy