Presenting a new method for retrieving images using their semantic classification

Abstract

With the growth and proliferation of digital technologies in recent years, the number of digital images has dramatically increased. ease of use of these technologies and the availability of effective tools for storing and transferring images has caused huge volumes of images to be made available to the general public. but the increasing number of images has caused problems, including the most important of which can be cited as a special image among the sums of images. Finding this particular image manual and looking at individual images is time - consuming, boring, and even impossible in some cases. Therefore, the need for an image recovery system that can extract desired image from the images is more than ever. the most important challenge in this approach is achieving an efficient way to investigate the similarity of images to each other and retrieve them. today, due to increasing importance of retrieving digital images, the need for a system that will perform the recovery process faster and more accurately. so the existence of an intelligent system that can achieve this will be necessary. in this thesis, a new method for retrieving images using their semantic classification has been presented. in the presented method, the image features were extracted using Descriptors (Resistant features against scale change)and LBP (local binary pattern)in terms of word packet method. finally, different investigations led to the creation of a novel method by combining Descriptors and LBP in a closed form of image vocabulary for image retrieval. the performance of the proposed method was measured using the svm classifier (support vector machine)with gaussian (support vector machine)on the famous OT database which consists of 2688 images in 8 different categories. the results showed that the proposed method has a high performance in retrieval of images using their semantic classification.