Visual Cognitive Attention Based Bag-of-Words Image Representation for Object Discovery

Abstract

In this paper, we proposed a Bag-of-Words image representation method in-spired by visual attention by applying computational visual attention technology to the representation of images, thus to boost the performance of the object discovery. First, a computational visual attention model was built on the real eye tracking data. With this attention model, we can find the most salient regions from the image, and then representing the image by emphasizes the visual words in these regions. Thus, we can get a Bag of Words image representation with more discriminative power, reducing the confusion intruded by the background on the images. Beyond discovering the objects from the images, with the guidance of the visual attention model, we are also able to find their locations. The experiment was carried out to verify the effectiveness of the proposed method. The experimental results showed that the proposed method improves the performance of the object discovery algorithm.