Eye Detection in Facial Image by Genetic Algorithm Driven Deformable Template Matching

As face recognition and facial feature based human computer interaction have become the subjects of intense focus in recent decades, facial feature extraction has emerged as a challenging task in the field of computer vision. Eye is said to be the most salient feature on face because of its versatility of appearance and expression variety. Various eye detection schemes are proposed in the literature but most of them require massive mathematical processing which is a barrier against real-time implementation. A novel approach for eye detection is proposed in this paper which exploits the flexibility of deformable template and uses genetic algorithm to match the template for eye detection. Implementation of genetic algorithm reduces the time required for template matching than conventional template matching algorithm. Moreover the method does not require any prior knowledge about eye geometry or potential eye location tags on facial image. Experimental results show that the proposed scheme can easily be implemented in real-time as it can detect eye in few genetic epochs. The method was tested on ORL face image database which contains 400 images grouped into 40 persons having 10 different expressions each. The detection accuracy was 87.2%.