An improvement of face detection algorithm for color photos

The problem of automatic detection of regions of interest (ROI) on color photos is considered. The efficiency of Viola-Jones face detector implemented in Intel OpenCV library is analyzed. The image size dependence of the algorithm characteristics is studied. It is shown that there are 2 significant drawbacks of this face detection algorithm as applied to the task mentioned. Some modifications are made for elimination of these drawbacks. To reduce false positives it is proposed to modify the algorithm with color segmentation and human skin tones analysis. To reduce the processing time the algorithm of downsizing preprocessing is proposed.