Content-Based Image Classification and Retrieval using Support Vector Machine

Abstract

Image search is more efficient for managing a wide range of image databases. Content-based image retrieval (CBIR) is one of the image retrieval techniques in which users use the visual characteristics of images such as color, shape and texture, etc. It permits the end user to give a query image in order to retrieve the image stored in the database based on the similarity to the query image. The system extracts the features of the query image, searches the database for images with similar features, and exhibits relevant images to the user in order of similarity to the query. Many CBIR systems have been developed to compare, analyze, and search images based on one or more of these features. This system is implemented as an image retrieval system combining visual content features and a support vector machine (SVM) classification. First, the system extracts the features of images from dataset with color autocorrelogram, color moment and gabor wavelet for the training phrase. When the user input query image, the system extracts features with these feature extraction methods in the testing phrase. And then, the system applies support vector machine (SVM) classifier to classify the image. After that, the system compares feature vectors between the query image and image dataset. Finally, the system retrieves the relevant image with query image. The applied system uses Wang dataset for the purpose of training and testing the system. And other 100 images that are not from dataset is also used for testing system. The overall accuracy of the system is over 80% for all classes. The system is implemented with MATLAB programming language on window platform.