Title :
Improving the accuracy of the Euclidean distance classifier
Author :
Amadasun, M. ; King, Robert A.R.
Author_Institution :
Imperial Coll., London, UK
Abstract :
The Euclidean distance (ED) classifier has the advantage of simplicity in design and fast computational speed, but has poor classification accuracy. Using a new feature normalization technique and feature weighting, a substantial increase in accuracy is obtained with no significant increase in computational cost or complexity of design.
Keywords :
pattern recognition; picture processing; classification accuracy; euclidean distance classifier; feature normalization technique; feature weighting; Accuracy; Complexity theory; Computational efficiency; Euclidean distance; Manganese; Support vector machine classification; Weight measurement;
Journal_Title :
Electrical and Computer Engineering, Canadian Journal of
DOI :
10.1109/CJECE.1990.6592169
