Center-surround interaction with adaptive inhibition: a computational model for contour detection

Area of Science:

• Neuroscience
• Computational Vision
• Image Processing

Background:

• The non-classical receptive field (nCRF) of primary visual cortex (V1) neurons modulates responses within the classical receptive field (CRF).
• This nCRF modulation is primarily suppressive and plays a role in visual information processing, such as contour extraction.

Purpose of the Study:

• To present a novel two-scale contour extraction model inspired by V1 neuronal inhibitory interactions.
• To investigate the distinct roles of nCRF side and end subregions in contour detection.

Main Methods:

• Developed a model incorporating adaptive inhibition based on interactions between CRF and nCRF.
• Implemented a mechanism where end region inhibition strength varies with local features at multiple scales.
• Side inhibition strength is based on local features at a fine scale.

Main Results:

• The model effectively removes non-meaningful texture elements while extracting object contours.
• Demonstrated superior performance in contour detection compared to other inhibition-based models.
• Showcased the adaptive mechanism's ability to differentiate contour locations from stochastic textures.

Conclusions:

• The proposed model provides a better understanding of nCRF roles in visual processing.
• The adaptive mechanism significantly enhances contour extraction from complex backgrounds.
• The model has potential applications in computer vision and pattern recognition.