Scientists from the visual systems lab at the Russian Institute for Information Transmission Problems (IITP RAS) have won first place in the international Illumination Estimation Challenge, beating the Google team, which took second place, as well as teams from China and Spain. The competition was one of the preparatory events for the International Symposium on Image and Signal Processing and Analysis (ISPA 2019).
The human visual system is capable of recognizing the colors of objects regardless of the illumination of the scene. Computer vision systems are confronted with a similar challenge. Essentially, they need to estimate the illumination and chroma. The participants were to develop a computer vision algorithm that would determine the color of the light source for several hundred photographs taken in various places and conditions using the same digital camera.
The IITP RAS team has developed an algorithm that best coped with the task, showing minimal deviations on the set of test images.
“Such algorithms are a hot topic not only in the scientific community, but also in industry. All known industrial giants are busy developing mobile phone cameras with high color rendering. Our solution can be used in next-generation cellphones to preset the white balance,” one of the algorithm developers said.