Weighting in the Template Matching
Jan Kalina · University of Lodz Repository (University of Łódź) · 2008
In the area of image analysis of templates, we have proposed and implemented a method for locating landmarks in 2D-images of faces. It uses the weighted correlation coefficient as a similarity measure between the template and the image. The weights are selected in the optimal way to improve the discrimination of parts of the image which correspond to the template from those which do not. The method does not use specific properties of faces.