| Human face detection and face organ feature location are both importantsteps in automatic visual interpretation and human face recognition. Facedetection is the very first and very important step of face recognition system,and it can also be implemented in video conferencing, crowd surveillance, andintelligent human computer interface. Face organ location, especially eyefeature location is an important step in a human face recognition system. Eyelocation is a crucial step towards automatic face recognition and man-machineinteraction such as gaze tracking and expression recognition due to the factthat these face related applications require normalizing faces, measuring therelative positions or extracting features according to eye positions.In video surveillance and other related field, it is difficult though usefulto detect low resolution human faces. For these small objects, as it is difficultto find the precise location of face organs, we need to use facesuper-resolution algorithm to improve our work.In this paper, a robust system which focuses on face detection and organlocation is introduced. We spend a lot of approaches on training samples forour face detector; we improve cascade-structured classifier of the face detector,and integrate human face super-resolution into our system to help processinglow-resolution faces. We use template matching, PCA combined with RealAdaBoost algorithm to get the precise location of eye feature, and the locationof other face organs is incorporated in the similar framework.The experimental results demonstrate that our system has goodperformance on actual images. |
