Hybrid Content Based Image Retrieval System Using Exhaustive Feature Set Processing by Multi Objective Optimization

摘要

In real world digital image processing, Content Based Image Retrieval (CBIR) is an emerging concept, it is an image retrieval framework based on some visual semantic features like color, texture and shape to search relevant images from multiple image sources. CBIR consist single query or multiple queries to semantic data or same object for multiple class labels with reference to multi query images. In retrieval of query image comparison from multiple image sources may cause optimization problem in image retrieval because of ambiguity in image search. To solve optimization problem for efficient query image retrieval from multiple image archives, in this study, we propose and develop hybrid content based image retrieval system. This system consist pareto optimized method (for removing non-dominated features) and Genetic Algorithm (GA) for effective content based image retrieval. Integration of contourlet features with color, texture and shape is considered to build exhaustive feature set for all images. In our approach Pareto optimal solution consist multi-objective selection functions to process image retrieval from multiple image sources for individual feature selection procedure with query image evaluation for image retrieval. This approach gives better performance than traditional approaches in query image retrieval from multiple image archives. Our experimental results show effectiveness of the proposed method in retrieving all images from image sources with improvement sorting the individual images in searching Pareto solutions in multi-objective optimization problem. Furthermore, this approach gives better accuracy with respect to precision, recall and time efficiency in image retrieval from multiple image sources based on extensive visual features.