A Method of Colour-Histogram Matching for Nigerian Paper Currency Notes Classification.

In this paper a new algorithm for classification of three Nigerian paper currency notes, namely 200, 500, and 1000 Naira (N) denominations is presented. The work examines the effectiveness of using only colour histograms to differentiate between the classes or denominations of the three Nigerian paper currency notes. The bin-heights of the histograms of the HSI component images for the paper currencies are used as features while a rule-based classifier designed to take advantage of the changes or variations in the histogram patterns is used to classify the paper currencies into the right denomination class. The algorithm involves the utilization of a simple and effective comparison strategy as opposed to the existing, too-rigid metrics for histogram-comparison used by other authors for color indexing in content-based image retrieval systems. Over a testing data-set of 300 samples, the algorithm achieved an average classification accuracy of 98.66%, and classification accuracies of 100%, 99% and 97% for the N=200, N=500 and N=1000 denominations, respectively. The proposed algorithm does not require extensive preprocessing of the paper-currency images and as such is fast in implementation.