点間の近接と連続性評価に基づく投票処理による主観的曲線の検出

抄録

This paper proposes a method to extract subjective visual curves in an image including discontinuous curves and noisy points based on the proximity and the continuity in the perceptive grouping factors. A pair of points is selected as the starting point and the goal point, and some points are connected sequentially from the starting point by evaluating the proximity and the continuity between neighboring points. The global form of the curve that is composed by the lines between the connected points is also evaluated by the proximity and the continuity, and the evaluated value is voted into the voting space. After the voting process about all pairs of points is completed, the addresses that have the large voted values are extracted as the components of curves. Experimental results show that the curves were extracted that had the similar forms to be perceived by the visual sense in noisy images.