Automatic red-eye detection and correction algorithm

According to the artifact known as red-eye of photographs taken at night,an automatic red-eye detection and correction method was presented.Firstly used AdaBoost algorithm to detect faces,could reduce the red-eyes' searching area.Then found the possible red-eye regions by the property of CIELAB color space,obtained ultimate red-eye location to filter the possible red-eye regions by constraints.The human eyes seem more natural after correcting the pixels of the ture red-eye regions and smoothing the pixels on the boundary.Experimental results show the proposed algorithm can detect and correct the red-eyes automatically,and it is very robust to typical red-eyes and red-eyes which are yellow or dark.