Face detection algorithm based on a cascade of ensembles of decision trees

Abstract

Face detection algorithm based on a cascade of ensembles of decision trees (CEDT) is presented. The new approach allows detecting faces other than the front position through the use of multiple classifiers. Each classifier is trained for a specific range of angles of the rotation head. The results showed a high rate of productivity for CEDT on images with standard size. The algorithm increases the area under the ROC-curve of 13% compared to a standard Viola-Jones face detection algorithm. To test the applicability of the algorithm in the real world have been conducted research on a robustness. Robustness research shown that the algorithm based on the CEDT show that Gaussian noise, impulsive “salt-and-pepper” noise exert a strong influence on the algorithm (in the worst case decrease in the area under the ROC-curve of 21.2% with a decrease in PSNR metric to 17.99 dB). At the same time blurring, JPEG-compression and JPEG2000 algorithms distortion have little effect on the proposed face detection algorithm (reduction of the area under the ROC-curve by 3.5% while reducing PSNR metric to 21.58 dB).