局部空间-时间显著性的表达过程蕴涵对空间显著性的表达

建立的基于独立成分分析进行特征提取器的学习,进而模拟人类对视觉信息的局部空间-时间显著性的表达过程的计算模型,蕴涵了对局部空间显著性的有效表达;从而当据其设计为机器人主动视觉的注意机制的视线控制而用的算法,可以同一个处理过程适用于动态场景或静态场景的情形。对获自动态场景的特征提取器的频率特性的分析,揭示了此种内在的包含性的成因;算法的实验结果显示了局部空间-时间显著性表达过程对静态场景的局部空间显著性可有效表达。

The Process of Locally Spatiotemporal Saliency Representation Revealing Its Validity for Pure Spatial Information

By adopting independent component analysis method to learn feature extractors, the developed computational model for the process of local saliency representation in human vision, intrinsically embeds the process for representing locally spatial salience in static visual information into the one of locally spatiotemporal representation. Therefore, algorithms based on the model for controlling the gaze of one robotic active vision is valid in both the case of dynamic scene and the one of the static. Frequency analysis on feature extractors obtained from dynamic scene reveals the cause of such intrinsical includedness. Experimental results show the validity of using locally spatiotemporal representation to measure the locally spatial salience of pure spatial information.