High Resolution Segmentation with a Time-of-Flight 3D-Camera using the Example of a Lecture Scene

Time-of-flight principle range cameras are becoming more and more available. They promise to make the 3D reconstruction of scenes easier, avoiding the practical issues resulting from 3D imaging techniques based on triangulation or disparity estimation. Their low resolution and frame rates as well as their low signal-to-noise ratio, however, kept many researchers from seriously considering this technology for segmentation purposes. The following article uses a practical application as an example to present both, an analysis of the problems resulting from the use of time-of-flight principle 3D cameras and a solution to deal with these issues. A lecturer is extracted out of a video that shows him in front of an electronic chalkboard. The extracted instructor can later be laid over the board semitransparently allowing the students to look through him in a later video replay. Using the 3D Time-of-Flight camera technology improves the robustness of the segmentation and lowers its computational complexity.