Real-time color tone detection on video based on the fuzzy integral

Color tone detection accomplishes the modelization of a color cluster for a set of pixels that present a hue similar to a particular one, which is being detected. Image pixels can be classified according to their membership to the particular color class through such cluster modelization. Such approaches can be employed in different computer vision application fields. Nevertheless few proposals in the literature report on the realtime detection of color clusters. This communication deals with the real-time implementation of the fuzzy integral, which is employed in a color cluster detection methodology. The resulting technique is applied to skin color detection and sport playground detection on video sequences. The method presented herein reaches a performance comparable to state-of-art methodologies with a very low computational cost, making real time color cluster detection possible. Both aspects, performance and computational cost, are commented herein and results are given in both standard benchmark databases and real-world application sequences.