Face detection in an image

摘要

In a system to detect a face in an image eg in an observer tracking display, candidate face regions are first identified S2 and then those candidate face regions are analysed for a characteristic indicative of a facial feature S4. Candidate face regions can be located by identifying areas of constant colour saturation in a reduced resolution version of the image. The analysing step may comprise selecting a single colour component, forming a vertical integral projection profile and detecting an omega shape in the profile characteristic of an eye region of the face. The aim is to reduce the data processing required by only analysing candidate face regions for the specific facial feature.