Elsevier

Neurocomputing

Volumes 58–60, June 2004, Pages 843-848
Neurocomputing

Determination of border ownership based on the surround context of contrast

https://doi.org/10.1016/j.neucom.2004.01.136Get rights and content

Abstract

We investigate the neural mechanisms for border-ownership (BO) determination, specifically whether the determination of BO is plausible from the contrast configuration within a certain range that extends beyond the classical receptive fields. The relevance of the contrast is suggested since the majority of BO-selective neurons in V2 and V4 show co-selectivity to the contrast. We hypothesize that the spatial structure of surrounding inhibition/excitation recently revealed in V1, or the similar structure in V2, is a key to integrate surrounding contrast to determine BO. The model reproduces a range of the neuronal activities responding to complex figures including occlusion.

Introduction

Discrimination of figure from ground is the essence for the perception of surface that is the fundamental source for the recognition of shape, spatial structure and objects. Recent physiological studies [1], [8] have reported that neurons in monkey's V2 and V4 showed the selectivity to border-ownership (BO) that indicates the direction of figure (which side of the border owns the contour). About 80% of the BO-selective neurons were co-selective to the contrast polarity of a border. This strong contrast dependency led us to expect that local contrast surrounding the classical receptive fields (CRF) could be a basis for the determination of the direction of figure.

We propose a network model for the determination of BO based on the surrounding contrast configuration that is determined following the spatial structure of the surrounding connections revealed physiologically. Although a number of models that utilized T-junctions and ownership junctions to determine BO have been proposed (e.g. [2]), these models may not be physiologically realistic since neurons selective to such junctions have not been reported. It should also be noticed that the latency of BO-selective neurons is several to 10ms [8]; thus complex processes such as the detection of ownership junction could not be involved. Our model determined BO solely from the surrounding contrast, and reproduced quantitatively the responses of the BO-selective neurons in V2 and V4. The simulations showed that a variety and asymmetry of inhibitory and excitatory connections are crucial for the determination of BO selectivity.

Section snippets

The model

We propose a network model for the determination of BO based on two principles that are physiologically plausible; (1) BO is determined exclusively from contrast context surrounding CRF, and (2) a variety in surrounding connections produces a variety of BO selectivity reported by Zhou et al. [8]. Physiological studies [3], [7] have shown that surrounding inhibitory connections beyond CRF are mostly asymmetric around the CRF, and that the extent of the suppressive region is larger than the

Simulation results

We carried out the simulations of the network model in order to investigate whether in fact the model reproduces the responses of BO-selective neurons in V2 and V4. The bar graphs in Fig. 2 show three examples of the results that indicate the responses of the model neurons (black bars) together with the corresponding neuronal responses re-plotted from [8] Zhou et al. (white bars). Cell a (Fig. 2A) shows the selectivity to left and light-dark. Cell b (Fig. 2B) responds strongly to stimuli with a

Discussions

The simulation study showed that (1) the context of the contrast surrounding the CRF is capable of determining BO without T-junction or ownership-junction detectors, and (2) a variety in asymmetric surrounding connections reproduces a range of BO-selectivity. The present model has excitatory connections among the cells with the same preferred orientation, and inhibitory connections among all cells regardless of the preferred orientation. It has been suggested that the function of surrounding

Haruka Nishimura took her BS and MS in Information Science at the University of Tsukuba, Japan, and is continuing the study toward a Ph.D. at the Graduate School of Systems and Information Engineering.

References (8)

There are more references available in the full text version of this article.

Cited by (0)

Haruka Nishimura took her BS and MS in Information Science at the University of Tsukuba, Japan, and is continuing the study toward a Ph.D. at the Graduate School of Systems and Information Engineering.

Ko Sakai received his BEng and MEng from the University of Electro-communications, Tokyo, and a Ph.D. in Bioengineering from the University of Pennsylvania, Philadelphia, PA. He is currently an associate professor at the University of Tsukuba, working on modeling visual perception.

View full text