Medial Axis for 3D Shape Representation
Cortical representation of shape is a crucial problem in vision science. Recent physiological studies on monkeys have reported that neurons in the primary visual cortex (V1) represent 2D shape by Medial Axis (MA). Physiology has also shown that a set of smooth surfaces represents 3D shape in a higher stage (IT). Based on the physiological evidence, we propose that monocular retinal images yield 2D-MAs that represent 2D-surfaces in V1, and the 2D-MAs are fused to yield 3D-MA that represents 3D-surfaces in IT. To investigate this hypothesis, we developed a computational model based on the physiological constraints, and evaluated its power on the shape encoding. The model represented a variety of 3D-shapes including natural shapes, with reconstruction errors of around 0.2 regardless of the shape complexity. The results support the visual system encodes monocular 2D-MAs in V1 and fuses them into 3D-MA in IT so that 3D-shape is represented by smooth surfaces.
KeywordsMedial axis MA Shape representation 3D Binocular disparity
Unable to display preview. Download preview PDF.
- 3.Hatori, Y., Sakai, K.: Robust Detection of Medial-Axis by Onset Synchronization of Border-Ownership Selective Cells and Shape Reconstruction from Its Medial-Axis. In: Köppen, M., Kasabov, N., Coghill, G. (eds.) ICONIP 2008. LNCS, vol. 5506, pp. 301–309. Springer, Heidelberg (2009)CrossRefGoogle Scholar