Pulse Coupled Neural Network based topological properties applied in attention saliency detection

Topological properties having priority and invariance play an important part in cognition. This paper introduces a novel attention selection model of Pulse Coupled Neural Network (PCNN)-based topological properties and quaternion. In our model, using Unit-linking PCNN hole-filter expresses the connectivity, an important topological property, in attention selection. Using this novel model can obtain spatio-temporal saliency maps from the phase spectrum of a quaternion image or a video´s hypercomplex Fourier transform. The experimental results show that this approach reflects the real attention with more accuracy than Phase spectrum of Quaternion Fourier Transform (PQFT) method.