DEWA: A Multiaspect Approach for Multiple Face Detection in Complex Scene Digital Image
Hadi Setiawan, Adang Suwandi Ahmad, Iping Supriana Suwardi, Farid Wazdi · ITB Journal of Information and Communication Technology · 2007
A new approach for detecting faces in a digital image with unconstrained background has been developed.The approach is composed of three phases: segmentation phase, filtering phase and localization phase.In the segmentation phase, we utilized both training and non-training methods, which are implemented in user selectable color space.In the filtering phase, Minkowski addition-based objects removal has been used for image cleaning.In the last phase, an image processing method and a data mining method are employed for grouping and localizing objects, combined with geometric-based image analysis.Several experiments have been conducted using our special face database that consists of simple objects and complex objects.The experiment results demonstrated that the detection accuracy is around 90% and the detection speed is less than 1 second in average.