Elimination of distorted images using the blur estimation at the automatic registration of meeting participants

Abstract

The methods for estimation of blur and other quality metrics of digital images are discussed. The classification of modern methods of blur estimation used for real-time systems is presented. Several methods of image patch segmentation were applied for enhancement of the processing speed and reliability of image quality assessment. The proposed method of preliminary extraction of face area on the image and estimation of its blur was successfully used for elimination of distorted images before the face recognition stage in a system of automatic registration of participants in a smart meeting room. © 2014 Springer International Publishing.