中文摘要：

为提高频域视觉注意模型的生物可信性和有效性，在综合分析现有模型的基础上，通过改进超复数傅里叶变换（hypercomplex Fourier transform，HFT）模型使用的颜色空间、四元数图像虚部系数以及最终显著图的产生方法，提出政进HFT模型。在Bruce和Judd两个数据集上对改进模型进行评估，实验结果表明，该模型显著性检测视觉结果与人类视觉更加接近，相较HFT，受试者接受曲线下的面积在两个数据集上分别提高了0．97％和6．33％。

英文摘要：

Aiming at increasing the biological plausibility and effectiveness of hypercomplex Fourier transform （HFT） model, on the basis of comprehensive analysis of existing frequency-domain visual attention models, an improved HFT model was proposed by employing suitable color space, imaginary coefficients of its quaternion image and the method of generating the final saliency map. The improved model was evaluated on Bruce and Judd image datasets, and the experimental results show that the saliency maps achieved by the improved model are much closer to human visual system （HVS） and the areas under the receiver operating characteristic curve （AUC） of the improved HFT model on the two databases increase 0.97% and 6.33%, respectively.