Because of the primacy of human subjects in digital images, much work has been done to find and identify them. Initial face detection systems concentrated on frontal, upright faces. Recently, multi-pose detectors have appeared, but suffer performance and speed penalties. Here we study solutions to the problem of detection invariance faced with in-plane rotation of faces. Algorithms based on integral projections and block averages estimate face orientation correctly within ±10° in about 95% of cases, and are fast enough to work in near real-time systems. © Springer-Verlag 2004.
CITATION STYLE
Nicponski, H. (2004). Understanding in-plane face rotations using integral projections. Lecture Notes in Computer Science (Including Subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), 3212, 633–642. https://doi.org/10.1007/978-3-540-30126-4_77
Mendeley helps you to discover research relevant for your work.