Abstract
Images can evoke emotions in the viewers, but are rarely indexed based on their emotional contents because emotions are subjective. Given the ubiquitous use of images, it would therefore be desirable to have an automated classification system for recognition of human emotions induced by the visual content of images (REVC), which can index images based on the emotions that they may induce. The focus of research in REVC systems has been on identification of features that evoke different emotions and their strength. These features are then used to develop REVC systems that can accurately predict emotions, which a given image may evoke. This paper investigates the performance of a color moment-based REVC system, by identifying the suitable color moments, and measuring their effectiveness for emotion recognition. Using a combination of these features, a new REVC system is developed. The proposed method outperforms a commonly used benchmark that relies on more than 100 features to predict emotional content of the image, signifying the fact that using combination of multiple features does not necessarily improve the system performance.

This publication has 29 references indexed in Scilit: