A bio-inspired contour detection model using multiple cues inhibition in primary visual cortex

• Published in: Multimedia tools and applications 2022-03, Vol.81 (8), p.11027-11048
• Main Authors: Lin, Chuan, Wen, Ze-Qi, Xu, Gui-Li, Cao, Yi-Jun, Pan, Yong-Cai
• Format: Article
• Language: English
• Subjects: Biomimetics; Computer Communication Networks; Computer Science; Contours; Data Structures and Information Theory; Energy distribution; Methods; Multimedia; Multimedia Information Systems; Neurons; Special Purpose and Application-Based Systems; Visual tasks
• ISSN: 1380-7501; 1573-7721
• DOI: 10.1007/s11042-022-12356-7

The human visual system has efficient architecture for information reception and integration for effectively performing visual tasks like detecting contours. Physiological evidence has shown that most neuronal responses in the classical receptive field (CRF) of the primary visual cortex are modulated, generally suppressed by the non-CRF surround. These center-surround interactions are thought to inhibit or facilitate responses to edges according to other similar edges in the surroundings, which is useful for suppressing textures and enhancing contours. A biologically motivated model with subfield-based inhibition is proposed in this paper to improve the performance of perceptually salient contour detection relative to the existing single-neuron based inhibition model. A novel subfield based inhibition framework is presented, where the inhibition terms are combined with center-surround and surround-surround differences using multiple cues, including orientation based energy distribution and directional saliency within regions. Extensive experimental evaluation demonstrates that the proposed method outperforms most of competing methods, especially biological motivated ones.