Automatic ROI Detection and Evaluation in Video Sequences Based on Human Interest

Abstract

This paper describes an approach for automatic detection and evaluation of Region of Interest (ROI) based on human psychology of interest and relevance. It is very challenging to determine the cognitive boundary of the scene in real time due to its complexity in decision making. We have proposed a novel method to solve this problem by using human cognitive processes of attention, interest generation and relevancy. Our method successfully determines this cognitive boundary as ROI which is significant by its information content based on interest and relevancy. The effectiveness of ROI detection is checked with Eye tracker system. The highest detection accuracy found is 81.25% which is higher than the existing method. The ROI evaluation is compared with evaluation by human observers. The subjective correlation is found 85% with human evaluation. The experimental results show that our method is imitating human decision making for ROI detection and evaluation.