Analysis of Distance Measures in Content based Image Retrieval
Content predicated image retrieval (CBIR) provides an efficacious way to probe the images from the databases. The feature extraction and homogeneous attribute measures are the two key parameters for retrieval performance. A homogeneous attribute measure plays a paramount role in image retrieval. This paper compares six different distance metrics such as Euclidean, Manhattan, Canberra, Bray-Curtis, Square chord, Square chi-squared distances to find the best kindred attribute measure for image retrieval. Utilizing pyramid structured wavelet decomposition, energy levels are calculated. These energy levels are compared by calculating distance between query image and database images utilizing above mentioned seven different kindred attribute metrics. A sizably voluminous image database from Brodatz album is utilized for retrieval purport. Experimental results shows the preponderating of Canberra, Bray-Curtis, Square chord, and Square Chi-squared distances over the conventional Euclidean and Manhattan distances.
