Quality Assessment and Simulative Performance Measures of Content Based Image Retrieval System

Content based image retrieval (CBIR) from large resources has become a dominant research field and found wide interest nowadays in many applications. In this thesis work, we design and implement a content based image retrieval system that uses color and texture as visual features to describe the content of an image region. We use k-nearest neighbor (knn) and HSV color model to extract feature of images. Our contribution is to design ̳knn‘ feature vectors. We feature of images used to create database are color correlogram, color moments, gabor filter for mean amplitude & energy calculation, wavelet moments and histogram. The similarity measures used in this thesis to find similarity between query image & database images are Manhattan distance, Euclidean distance and Relative deviation. We segment the images into five classes consist images of Dinosaur, Bus, Beach, Flower and Sunset. We combine various features of images to construct ̳knn‘ feature vector. The proposed CBIR system using various feature of ̳knn‘ has the advantage of increasing the retrieval accuracy in form of Precision and Recall. The experimental evaluation of the system is based on a 250 color image database. From the experimental results, it is evident that developed system performs significantly better and faster compared with other existing systems like HSV. Keywords— CBIR; HSV; KNN; manhattan distance; euclidean distance.