H. Yoon, D. Kim, S.Y. Chi, and Y. Cho (Korea)
Head detecting, Moving camera, Ellipse fitting
In this paper, we describe an algorithm for head detection using the multi-cues combined by gray, edge and color features. This algorithm is used for robot control that has additional limits as moving camera, moving objects, long distance objects, various face orientations, and unfixed illuminations. In this circumstance, we can use only one assumption that head contours have an ellipse shape. In order to separate from background or body to head contours, we use three basic features gray, color and edge and combine them. Skin color cue is very useful when image stream have frontal and large face and have no background objects similar with skin color. Gray cue is also important when captured faces have low gray level than background objects. Edge cue is helpful when captured background objects have similar gray level and color with head but discriminate by edges. Since these three methods have roughly orthogonal failure results, they serve to complement each other. The results of this algorithm provide the robustness for head rotation, illumination changing, and variable head sizes. Also, it is possible to real time processing.
Important Links:
Go Back
