Abstract
For security requirements, a number of studies has been reported on personal identification using facial images. Most of them use facial images taken from one&primes front or side(s). The angel of the face to a camera influences the accuracy of processed results. Therefore, it is important to detect the front view of facial images correctly. In this paper, we define feature quantities and show their effectiveness in detecting specifically the front view of facial images in the horizontal and vertical directions. Our analytical results demonstrate that we can make good use of the feature quantities for the accurate and automatic detection of the frontal face direction.
