Enhanced Image Classification with Feature Level Fusion of Niblack Thresholding and Thepade’s Sorted N-ary Block Truncation Coding using Ensemble of Machine Learning Algorithms

Abstract

The paper portrays novel enhanced image classification approach with fusion of Machine Learning Algorithms at Feature Level as well as Decision Level with help of Niblack Thresholding and Thepade’s Sorted N-ary Block Truncation Coding. The proposed fusion based image classification method is experimented with help of a database with total one thousand image samples covering ten assorted image categories with 100 images per category. Classification Accuracy is taken into account for the performance evaluation purpose of existing and the proposed Image Classification Technique. The results of experimental analysis explicitly reveal the performance improvement with proposed TSnBTC than Niblack thresholding, also the fusion of these two methods reveal further better performance with several Classifiers proving the worth of proposed fusion based image classification technique. Overall the higher classification accuracy is given by Random Forest immediately followed by ensemble of Random Forest with SVM.