Applications of Hybrid Machine Learning for Improved Content Based Image Classification

Content Based Image Classification (CBIC) has scaled heights due to efficient applications of machine learning algorithms. Information identification using remotely sensed satellite images, medical diagnosis with image data, event detection in real time, catastrophe detection with images and many more are few instances of the immense possibilities achieved using machine learning for CBIC. The success of any classification algorithm has its pivotal contribution from the feature extraction technique adopted during descriptor designing from images. The authors have attempted to implement the hybrid approach of machine learning with two different feature extraction techniques. The work has explored both the early fusion and late fusion to experiment the possibilities of enhanced classification decision. The approaches are contrasted against each other for the measure of efficiency. The observation divulges superiority of the fusion based techniques over the individual techniques in terms of classification performances. Comparison with benchmarked techniques has also revealed encouraging results for the proposed technique.