中文摘要：

人类视觉系统通过选择性视觉注意机制来对场景中位于重要位置的视觉内容进行动态的序列采样,进而获取必要的视觉信息.本文系统地总结了计算化注意模型和显著度计算领域的研究现状.通过在YORK-120和MIT-1003这两个国际标准数据库上进行的人眼视点预测实验,本文对20种计算模型的实际性能进行了详细的评估和对比.结果表明,基于统计的模型要比其它的方法更容易获取较好的预测结果.

英文摘要：

Human vision system acquires essential information from the environment by sequentially sampling visual con- tents at important locations under the control of selective visual attention mechanism. This paper systematically investigates the state - of - the - arts of computational visual attention modeling and saliency computation. And the paper compares the performance of 20 state - of - art models via dense eye - fixation prediction experiments conducted on YORK - 120 and MIT - 1003 datasets. The results show that statistical models tend to achieve better predictions against the rest approaches.