Ice-hockey puck detection and tracking for video highlighting

Detection and tracking of ice-hockey pucks are of interest to sports industry as it allows puck highlighting during broadcast. This paper presents the first attempt at using image processing to detect and track ice-hockey pucks. The technique developed consists of using three pieces of information in ice-hockey video frames. A combination of adaptive gray-level thresholding, adaptive temporal thresholding, and trajectory constraints is used to detect and track pucks. The analysis performed indicates the difficulties of this detection problem when using ice-hockey broadcast video sequences. A comparison of the developed technique is provided with table tennis broadcast video sequences to demonstrate its effectiveness.