Abstract
Red-eye and flash-eye defects in still photography continue to cause problems for digital imaging devices. New variants of flash-eye defects have appeared as cameras and cameras sub-systems get smaller in size. This paper reviews a wide range of the more recently patented techniques that have been applied to solve various problems associated with red and non-red flash-eye, particularly solutions which are implemented in the camera. The use of in-camera metadata, multiple images of the same scene and adaptation of image acquisition conditions to improve red-eye detection and correction are explained. Four principle categories of flash-eye defect are identified and techniques are outlined to deal with each of these. The use of face, eye and scene metadata obtained within the camera is also considered. In particular data obtained regarding indoor/outdoor scenes, partial face regions, eye orientation and gaze and pupil size can all assist with some difficult cases of flash-eye defect. Some advanced techniques to improve the correction and reconstruction of eye regions are also considered.
Keywords: Embedded image processing, flash-eye defects, redeye detection, redeye correction, Connected Red Segments, Simple Filter Set, Eye-Glint for Detection, Camera red-eye filter, Camera Metadata, Image Filtering
14