Fig. 16
From: Understanding and simulating border ownership centered segmentation
Architecture Diagram for the Proposed Border Ownership-Centered Figure-Ground Organization (Segmentation) Model. The architecture integrates neuroscience-inspired elements such as global context awareness, functioning as ‘prior knowledge’, and active neurons. Inputs, including RGB images, disparities, optic flows, as well as outputs, such as contours, border ownership, and category selectivity, are represented using a channel-based format. This representation facilitates seamless integration across the visual processing pipeline, enabling efficient and scalable segmentation of contrast-, disparity-, illusory-, and contour-defined objects.
