Face Detection Using Boosted Cascade of Simple Feature

Abstract

Face detection from picture or capture is admired focus in biometrics study. A lot of public spaces typically include observation cameras for videotape capture and these cameras have their considerable value for safety intention. It is widely recognized that the face discovery includes a significant role in observation arrangement as it doesnt need the objects assistance. The actual compensation of face based recognition over other biometrics is individuality and acceptance. As a persons face is a dynamic thing that bears high degree of changeability in its look, it makes a computer vision to count face discovery as a tough problem. The main challenge in this field is the speed of detection and accuracy. This paper deals with estimating variety of face detection methods and provide a total solution for image based face detection with higher accurateness, an improved response rate as an initial step. The explanation is projected based on performed tests on various face rich databases. Initially, the segmentation of non-skin color pixels from the image is done by building a skin color model in chrominance space i.e. YCbCr. Later, in order to extract human face region, the mathematical morphological operators are used that removes noisy regions and fills the holes in skin-color region. Ultimately, to achieve face detection more accurately, the cascade classifier based on an AdaBoost algorithm is used to scan these face candidates.