Measurement and quantitative analysis of human characteristic on visual subjective contours

Pattern recognition technology has been widely applied to many application areas. However, the performance of pattern recognition system is nowhere near human visual ability. In order to develop the pattern recognition algorithms which match human visual ability, it is necessary to measure human visual characteristic of subjective contours. In this study, two psychological experiments of the subjective contours were conducted. In the experiment 1, we verified whether Shipley & Kellman finding of the subjective contours is suitable for dissymmetric figures. It was found that their finding can not explain the experimental results of the dissymmetric figures. We found that the subjective contours strength depend on the edges of interior angles. In the experiment 2, we changed the mode of experimental figure to make sure whether the results of the experiment 1 are still tenable, and the results confirmed that the conclusions of the experiment 1 are correct. Based on the experimental results, a new mathematical relationship between the strength of the subjective contours and the edge lengths is proposed. It is expected that the results reported in this paper are useful to develop new pattern recognition algorithms.