The Metric Histogram: A New and Efficient Approach for Content-based Image Retrieval

This paper presents the metric histogram, a new and efficient technique to capture the brightness feature of images, allowing faster retrieval of images based on their content. Histograms provide a fast way to chop down large subsets of images, but are difficult to be indexed in existing data access methods. The proposed metric histograms reduce the dimensionality of the feature vectors leading to faster and more flexible indexing and retrieval processes. A new metric distance function DM( ) to measure the dissimilarity between images through their metric histograms is also presented. This paper shows the improvements obtained using the metric histograms over the traditional ones, through experiments for answering similarity queries over two databases containing respectively 500 and 4,247 magnetic resonance medical images. The experiments performed showed that metric histograms are more than 10 times faster than the traditional approach of using histograms and keep the same recovering capacity.