Feature alignment techniques for pattern recognition
P. Kosir, R. DeWall · 2002
Feature alignment is a necessity for most pattern recognition systems. By definition, a parametric pattern recognition system attempts to classify an object based on a statistical model of the features from that object. These feature statistics are determined during the training process through multiple measurements of an object. In many cases, however, the starting location of these features varies from measurement to measurement. It is therefore important to align the features from one measurement to the same features from all other measurements, in order to obtain a proper statistical estimate of the features. Failure to do a good job at aligning the measurements will result in a corrupted statistical estimate of an object's features; hence, significantly degrade the performance of the pattern recognition system. A critical part of most pattern recognition systems is the feature alignment function. This paper will present same background information on pattern recognition and illustrate the importance of feature alignment. This paper will also discuss the use of the correlation function as a tool for feature alignment. Three different alignment techniques which use the correlation function will be described.>