Review of Image Memorability Prediction

Abstract

With the development of information technology and networks, images can be seen everywhere in life. When people browse magazines and the Internet, they come into contact with thousands of images, some of which remain in people's memories, while others are forgotten. Isola et al. first proposed the concept of image memorability and proved that image memorability is an intrinsic and stable attribute of images and can be shared among different viewers. Image memorability prediction has important research value and can be applied to education, advertising and other fields. In order to promote the research of image memorability prediction, the main theories and methods are summarized. Based on the comparison and analysis of the literatures related to image memorability prediction, this paper firstly reviews the proposal and quantification of the concept of image memorability. The features that affect the image memorability and the methods for extracting features are analyzed. The prediction of predictive image memorability is introduced in detail. The prediction is constructed based on Support Vector Machine and Convolution Neural Network and using different features that affect image memorability. Finally, the future work of image memorability is summarized and forecasted.