A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2010; you can also visit the original URL.
The file type is
A probabilistic, maximum aposteriori approach to finding landmarks in a face image is proposed, which provides a theoretical framework for template based landmarkers. One such landmarker, based on a likelihood ratio detector, is discussed in detail. Special attention is paid to training and implementation issues, in order to minimize storage and processing requirements. In particular a fast approximate singular value decomposition method is proposed to speed up the training process anddoi:10.4304/jmm.5.3.276-289 fatcat:7vwdxgvp2rcenltpqa6erumzua