LG248: Feature normalization and likelihood-based similarity measures for image retrieval

Date: 
Friday, October 30, 2015 - 12:00
Speaker: 
Ahmad Algohary and Rakesh Shiradkar, PhD
Abstract: 
Distance measures like the Euclidean distance are used to measure similarity between images in content-based image retrieval. Such geometric measures implicitly assign more weighting to features with large ranges than those with small ranges. This paper discusses the e€ects of ®ve feature normalization methods on retrieval performance. We also describe two likelihood ratio-based similarity measures that perform signi®cantly better than the commonly used geometric approaches like the Lpmetrics.