Novel data mining based image classification with Bayes, Tree, Rule, Lazy and Function Classifiers using fractional row mean of Cosine, Sine and Walsh column transformed images

Important task in image database is to organize images into appropriate category using different features of images. Image classification is studied for many years. There are various techniques proposed to increase the accuracy of classification. In this paper a novel data mining based approach is proposed for content based image classification. Feature extraction and classification algorithms are two main steps in classification process. This paper proposes the use of orthogonal transform to generate the feature vector and to investigate effectiveness of different transforms (Cosine, Sine, and Walsh). Experimentation is carried on different sizes of feature vectors which are formed by taking fractional coefficients. Classification algorithm from different families such as Bayes (Naive Bayes and Bayes Net), Function (RBFNetwork and Simple Logistic), Lazy (IB1 and Kstar), Rule (Decision and Part) and Tree (BFTree, J48 Random Tree and Random Forest) are used for classification. Experimental results and its analysis have shown the Simple Logistic classifier with Walsh transform to be better for proposed data mining based image classification technique.