【摘要】 编组模型大多基于视知觉格式塔规则中接近率、连续率、共线率和相似率,没有考虑全局约束规则如对称率和平行率作用,编组结果缺乏视觉上完整性问题,对此该文提出一种全局约束的层次编组模型。选取显著性重要的小波变换变换边缘作为编组候选数据集,利用格式塔规则中对称率和平行率作为全局线索指导编组过程。不同类型的图像实验表明,模型在不损失视知觉合理性的同时减少了编组过程中的二义性,增强了编组效率和健壮性。更多还原

【Abstract】 Although numerous studies measure the strengths of perceptual organization cues for grouping contour, most ignore the roles of symmetry and parallel principles in grouping contour.A hierarchy grouping model is developed.Firstly, most saliency edges are extracted as input for subsequent contour grouping.Secondly, the symmetry measure as higher grouping cues which supervises the low level grouping process.Experiments on different types of images show that the model reduces ambiguities in the intermediate representation and have higher efficiency and robustness, while holding biological plausibility.更多还原