The Selective Attention for Identification Model (SAIM): Simulating Visual Search in Natural Colour Images

Abstract

We recently presented a computational model of object recognition and attention: the Selective Attention for Identification model (SAIM) [1,2,3,4,5,6,7]. SAIM was developed to model normal attention and attentional disorders by implementing translation-invariant object recognition in multiple object scenes. SAIM can simulate a wide range of experimental evidence on normal and disordered attention. In its earlier form, SAIM could only process black and white images. The present paper tackles this important shortcoming by extending SAIM with a biologically plausible feature extraction, using Gabor filters and coding colour information in HSV-colour space. With this extension SAIM proved able to select and recognize objects in natural multiple-object colour scenes. Moreover, this new version still mimicked human data on visual search tasks. These results stem from the competitive parallel interactions that characterize processing in SAIM.