Study of Viola Jones Face Detection on Color Image based on Skin Pigmentation Level

Abstract

Automatic face detection has been very complex and challenging research topic due to the complexity of faces’ characteristics that is not rigid object. There have been many works on proposing robust algorithm on image detection. Many researcher use Viola Jones algorithm as their initial point and benchmark. The Viola-Jones face detection itself is the most popular and recent applicable algorithm that has been developed since 2004 by Paul Jones from Microsoft R&D and its co-inventor, Michael J. Jones from Mitsubishi R&D. Many previous works present the study on the Viola Jones algorithm subject to frontal face with no consideration on the skin pigmentation level. This paper presents study on The Viola Jones performance on color image that consider skin pigmentation level. To indicate the skin pigmentation level, the L* element on CIELAB color space is used. The skin pigmentation level is clustered into dark skin, brown skin and fair skin. The simulation result show that the Viola Jones performance tends to decrease when the skin pigmentation getting high (dark skin). Some hypotheses test had been done to support the claim.