An image retrieval approach based on feature extraction and self-supervised learning

Abstract

Today, due to the development of technology and the advent of web 2.0 applications, different users prefer to do many of their personal tasks over the Internet. Due to the huge amount of information on the web, retrieving the appropriate information for each user has become a challenging task. Content-based image retrieval is one of the most important research fields in digital image processing domain, which searches the similar images to the target image by extracting visual content from the query image. In this regard, many studies have been conducted to increase the accuracy of image retrieval systems. However, due to the explosive growth of storage resources and the lack of a responsible system for image retrieval, it is still considered as one of the most attractive fields of research. In this paper, a method is proposed that extracts the appropriate features using a hybrid method, and then searches the images that are similar to the target image. In this way, self-supervised learning approach is utilized to provide the most similar images. Experimental results based on the Corel dataset show that the accuracy of the proposed method has increased compared to the other methods.