Content based image retrieval using novel gaussian fuzzy feed forward neural network

Abstract

quotThis thesis presents research work related to Content Based Image Retrieval with focus on medical images. This is an extension to search based on text query wherein images are applied as text query. This works extends to image based query. newline newlineThis study handles the problem related to the difficulty of handling high level image content from low level image features. Retrieval method based on relevance feedback require human interaction. newline newlineThe proposed approach uses fusion of texture and shape features. Texture features used were mean, variance, standard deviation, contrast, energy etc. whereas shape features used are area, perimeter, circularity, aspect ratio etc. Euclidean and Manhattan distance were used to extract relevant images from database. Successively artificial neural network was applied for image classification. newline newlineThe proposed approach with composite features was experimentally compared with existing approach on dataset of 250 images and was found effective and superior for medical classification. newline newline