Infrared Image Brightness Correction for TIR Object Tracking

Thermal imaging has been widely used in numerous applications of computer vision. However, in infrared videos, drastic change to bright or dark is a common phenomenon. 8bit thermal image suddenly turning bright or dark mainly caused by two reasons: change in light and dark pixel values distribution and the temperature change inside the infrared camera during photography. As a result, both the tracking performance and the visual effect are affected. To this end, we analyzed the problem from the principle of infrared imaging. We proposed a brightness correction algorithm based on the Kalman filter, which suppressed adjacent frames’ brightness flickers, thus improving both tracking performance and visual effect. Experiments on a standard tracking benchmark LSOTB-TIR demonstrate that our algorithm achieves better tracking performance on flickering sequences while not negatively influencing regular infrared videos.