Kinect v2 infrared images correction

Abstract

This article presents a novel correction filter for infrared images captured by Kinect v2 sensor. Intended application areas are described and the basic concept of proposed sensory subsystem as well. Preliminary tests of the sensor Kinect v2 in real conditions gave promising results, therefore in-depth analysis of its applicability was performed. In the framework of the analysis, a relation between infrared value and a distance of captured surface was evaluated for different colors of the surface. Based on that relation and on additional information about the depth of a pixel, a correction filter was created. The filter allows improving infrared image in order to increase the success rate probability to be able to detect specific features and key points by algorithms more easily. Outputs from the filter on real conditions data sets are also presented in the article.