A METHOD AND SYSTEM OF FILTERING A RED-EYE PHENOMENON FROM A DIGITAL IMAGE

摘要

A method of detecting and correcting an eye defect within an acquired digital image comprising a multiplicity of pixels indicative of color and forming various shapes within the image is disclosed. A probable size distribution of eye regions, if any, in the image is determined using image meta-data and human anthropometric data, the meta-data including information describing conditions under which the image was acquired. A candidate eye region in the image is detected and whether it falls within the size distribution is determined. In respect of a candidate eye region falling within the size distribution, the anthropometric data is used to search for the presence of at least one of a second eye, lips, nostrils and surrounding face. If the search is successful, a probability that the candidate eye defect region is an actual eye region is increased. The candidate eye region is confirmed as an actual eye region if the probability is above a certain threshold. If the confirmed eye region has a color representative of a red-eye defect, the red-eye defect is corrected.