Robson Barcellos,  Kátia Veloso Silva,  Osvaldo Severino Junior,  Adilson Gonzaga. 

The performance of CBIR systems is heavily dependent on the choice of the method used to generate feature vectors (FV). In this work a comparison of the performance of a CBIR system is made when three different methods are used to generate FVs; autocorrelograms, mixturegrams and perceptual color histograms. Also, two different image data bases are used to show how the nature of the images affects the results. Precision vs. Recall graphs are used to compare the performance.

http://www.lbd.dcc.ufmg.br/colecoes/wvc/2006/0056.pdf

Processando consulta.