Detecting Faces in Images: A Survey

This paper is a survey of methods for detecting faces in images, motivated by the fact that vision-based human-computer interaction tasks, including face recognition, face tracking, pose estimation, and expression recognition, often assume that faces have already been identified and localized. It defines the face detection problem as, given a single image, identifying all image regions that contain a face regardless of its 3D position, orientation, and lighting conditions. The authors stress that this is challenging because faces are non-rigid and vary greatly in size, shape, color, and texture.

The survey categorizes and evaluates the numerous techniques developed for single-image face detection and discusses related practical issues such as data collection, evaluation metrics, and benchmarking. After analyzing the algorithms and identifying their limitations, the authors conclude with several promising directions for future research. By organizing a fragmented field and establishing common evaluation considerations, the paper served as a foundational reference for building fully automated face-analysis systems.