Robust Face Detection through Eyes Localization using Dynamic Time Warping Algorithm

Detection of faces and facial patterns in static or video images is an important but challenging problem in computer vision, as faces may present in different scales, orientations, positions and poses in an uncontrolled background. To date, numerous approaches have been implemented for face and facial pattern detection [1-4]. These approaches have been grouped into four broad categories [2]. These are knowledge-based, feature based, appearance based and template based methods. In this chapter, we present a simple yet robust algorithm for template matching method for face detection through eyes localization by using Dynamic Time Warping (DTW) algorithm with different poses and variations. By using DTW, we overcome some of the main limitations of Classical template matching. We elaborate through discussion and experiments’ results that how our image processing strategy applied to DTW algorithm effectively increase the performance and detection rate. Using such an approach does not require translation of human knowledge about the face and facial patterns to computer representation, training data set and training mechanisms or face geometry, and does not require large numbers of model templates to handle different pose and variation changes shown by facial features. In the coming section, short study of Dynamic Time Warping algorithm (DTW) is presented. Proceeding to the next section we discuss the necessary steps for our image processing strategy. This is followed by presenting our approach of image partitioning, and in the next section we present dynamic time warping algorithm application in our work. Towards the end of this paper, results and discussion are put forward and conclusion is presented.