Size-based performance analysis of haar-features for detection of facial images from low resolution surveillance videos

Face detection has been a very prominent research field since past two decades. In most of the works done so far, haar-features have been the key components in the task of object detection and have been very instrumental owing to their simplicity, efficient computation, scalability and ability in the encoding of structural contents of facial images in the form of compact binary representation. The selection of the size of example images for the training of the haar-features decides the base size of the detector which consequently influences the overall performance of the face detection system. In this work, we have done the performance analysis of modified version of face detection algorithm with respect to the size of the haar-features based trained detectors and applied the concept to process facial images from low-resolution surveillance videos. It has been empirically concluded that hybridization of the classifiers from different base size results in better detection rate as compared to using haar-features trained from the single base size in detecting low-resolution faces.