A GMM Based Automatic Red Eye Detection Algorithm

In digital photography Red eye is caused mainly under low light conditions when a flash is used for capturing the image. Many of the commercial solutions that are available in the market require some user intervention. In this paper we present a method, which detects the presence of red eyes in Images without any user intervention and corrects it. The automatic solution is based on GMM (Gaussian Mixture Model) classifiers for the detection stage. The proposed method corrected about 81% of the red eye artifacts in a test image database of 500 images, resulting in very few wrong corrections. The computational complexity of running this algorithm on a consumer electronics device such as a digital camera or a mobile phone is also very less.