【摘要】 受人类视觉系统采用层次性信息处理机制预测感兴趣区域和进行快速目标检测的启发,提出了一种新的基于贝叶斯推理的层次性视觉注意模型。该模型利用来自腹侧通路的颜色特征作为自底向上的推理信号,利用来自背侧通路的方向位置信息作为自顶向下的推理信号,并将这些信息在贝叶斯框架下集成,通过推理得到显著图。用自然场景图像集进行了该模型与IT、GB等9种视觉注意模型的性能对比实验,结果表明该模型的显著图更接近于人类注视图,能够更好地模拟视觉注意。此外,在遥感图像上的实验表明,与性能接近的GB模型相比,该模型能更好地实现目标检测。更多还原

【Abstract】 Inspired by the human visual system’ hierarchical processing mechanism for prediction of interesting regions and quick detection of targets,a novel hierarchical visual attention model based on Bayesian inference is proposed.The model utilizes color feature information from the ventral pathway as the bottom-up inference cue,and uses orientation location information from the dorsal pathway as the top-down cues.All the cues are integrated in a Bayesian framework,and saliency maps are obtained by inference.The natural image experiments were conducted to compare the performance of the proposed model with that of the other nine state-of-the-art visual attention models IT,GB,AIM,ECSF,CA,SUN,SR,QB and FT,and the results show that the saliency maps created by the proposed model are much closer to the eye fixation data,and it can better simulate visual attention.Moreover,the results of the experiments on remote sensing dataset demonstrate that the proposed model can better detect targets compared to the GB model with the similar performance.更多还原