A novel approach for fast circlet transform: dynamic analysis of coefficients for circular shapes quantification

Abstract

Efficient detection of circular-shaped patterns in images remains a persistent challenge in computer vision applications. The commonly used algorithms have limitations in accurately defining circle parameters like size, quantity, and depth cues. This study proposes a trajectory dynamic analysis technique by probing the sequential evolution of the Fast Circle Transform coefficients. The transitional behavior of FCT coefficients reveals extremely distinctive convergence-divergence patterns as the circlet size varies, attributable to phase alignments. Rigorous evaluations involving simple circular patterns and imagery revealed that the proposed methodology exhibits excellent detection capability, similar to human visual perception. The novel recursive trajectory-based technique facilitates the detection of multiple circular features. Through the integration of concepts from different domains and the transition from static to dynamics-driven analytics, this research offers a methodology to address ongoing challenges in shape quantification and depth estimation. The expansion of its generalizability will broaden its applicability to various vision tasks.