Figure-Ground Segregation by a Population of V4 Cells A Computational Analysis on Distributed Representation

• Published in: Neural Information Processing pp. 617 - 622
• Main Authors: Hasuike, Masaharu, Ueno, Shuto, Minowa, Dai, Yamane, Yukako, Tamura, Hiroshi, Sakai, Ko
• Format: Book Chapter
• Language: English
• Published: Cham Springer International Publishing 10.11.2015
• Series: Lecture Notes in Computer Science
• Subjects: Electrophysiology; Figure ground segregation; Natural image; Population coding; Support vector machine; Visual cortex
• ISBN: 9783319265346; 3319265342
• ISSN: 0302-9743; 1611-3349
• DOI: 10.1007/978-3-319-26535-3_70

Figure-ground (FG) segregation is a crucial function of the intermediate-level vision. Physiological studies have suggested that monkey V4 cells are sensitive to FG organization, although their selectivity to FG and what they represent have not been clarified. We investigated computationally whether a subpopulation of cells in V4 represents FG. Specifically, we applied support vector machine to the responses of 1–100 cells to natural-image patches that were recorded from V4 of two macaque monkeys (Macaca fuscata) by multi-electrode arrays. Our results showed that the responses of a group of cells were capable of segregating FG while the responses of single-cells were not, suggesting a distributed representation of FG in V4.