||Implementing the Viola-Jones Face Detection Algorithm
||Jensen, Ole Helvig
||Larsen, Rasmus (Image Analysis and Computer Graphics, Department of Informatics and Mathematical Modeling, Technical University of Denmark, DTU, DK-2800 Kgs. Lyngby, Denmark)
||Technical University of Denmark, DTU, DK-2800 Kgs. Lyngby, Denmark
||The emergence of high resolution digital cameras for recording of still images and video streams
has had a significant impact on how communication and entertainment have developed during the
recent years. At the same time Moore’s law has made tremendous computing power readily
available that only some 20 years ago was reserved for high profile research establishments and
intelligence services. These two tendencies have respectively called for and fostered the advent of
unprecedented computationally heavy image processing algorithms. Algorithms that in turn have
allowed for new processing of existing image based material.
Parallel to this technological development the measures deployed in the protection against attacks
from the enemies of modernity calls for more surveillance of the public space. As a result of this
regrettable circumstance more video cameras are installed in airports, on stations and even on open
streets in major cities.
Whether or not the purpose is entertainment or dead serious surveillance, tasks like detection and
recognition of faces are solved using the same methods. Due to the varying and generally adverse
conditions under which images are recorded there is a call for algorithms capable of working in an
In 2004 an article by Paul Viola and Michael J. Jones titled “Robust Real-Time Face Detection”
was publish in the International Journal of Computer Vision. The algorithm presented in this article
has been so successful that today it is very close to being the de facto standard for solving face
detection tasks. This success is mainly attributed to the relative simplicity, the fast execution and
the remarkable performance of the algorithm.
This report documents all relevant aspects of the implementation of the Viola-Jones face detection
algorithm. The intended input for the face detection algorithm is any conceivable image containing
faces and the output is a list of face positions.
Creation date: 2008-09-19
Update date: 2008-09-19