A novel computation technique for single and multiface detection using Euclidean distance and principal component analysis (PCA)

Face detection system attracts huge attention in recent years due to it may improve security of surveillance systems. In developing a face detector system, there are sub problems ariseone of these sub problems is the low accuracy. Principal Component Analysis (PCA) is well known to be one of the methods for face recognition and detection, in where a threshold value has to be fixed in the Euclidean distance computation. Computing a fixed threshold for multi environment is very difficult which consequently leads to performance reduction. As such, this paper proposes a method which does not rely on threshold value but instead, merely relies on Euclidean distance between two subspaces. A standard database developed by Massachusetts Institute of Technology (MIT) Centre for Biological and Computation Learning (CBCL) is used to evaluate the proposed method. In the testing stage, real life images are used as well. Comparison results between the proposed method and the original method show that the proposed method can reduce the dimension until 60% and has a good competent accuracy (89.34%) for single and multiface detection although performs slower than normal PCA