3.9 Summary

In this chapter we have introduced several computational models of spatial bottom-up visual attention based on the background knowledge introduced in Chapter 2; these include the original baseline salience (BS) model, the BS model with motion feature, the BS variations, GBVS model, the AIM model, the DISC model, the SUN model and the model based on surprise theory. In early studies in this field, most of the computational models point towards completely simulating the visual structure in the human eyes and brain, such as in the BS models. During recent years, some statistical signal processing theories that are frequently applied to engineering areas have been introduced in computational visual attention models, and both biologically inspired ideas and signal processing methods are considered. The computational models based on signal processing theories appear more effective than purely biological models. All the models in this chapter are computable and programmable, so they are easy to implement in various real-world applications.