earticle

영어

This paper describes the use of histogram modification functions to improve the retrieval efficiency of the content-based image retrieval system based on bins approach. Four different functions explored in this paper used as histogram specification to modify the histograms are; histogram equalization (EQH), polynomial function (POLY), linear equations (LinearEQ1,2,3) and logarithmic function (LOG). Modified histograms are partitioned into 4 parts using centre of gravity (CG) to form 64 bins from 256 bins of R, G and B histograms. Bins holding the count of pixels falling in particular range of modified intensities divided into four parts using centre of gravity. Statistical properties are computed for the intensities possessed by the pixels counted into these 64 bins. These properties are representing the type of feature vector in the form of first four moments namely mean, standard deviation, skewness and kurtosis. All moments are computed for three-color intensities R, G and B separately. Based on color and moment different types of feature vector databases are prepared. Comparisons of query feature vector with feature vectors of database images is carried out by means of three similarity measures namely Euclidean distance (ED), absolute distance(AD)and cosine correlation distance(CD). Role of each modification function along with all types of feature vectors is evaluated and compared and presented using three parameters; Precision Recall Cross over Point (PRCP), Longest String(LS), Length of String to Retrieve all Relevant (LSRR).