A Clustering-Based Approach for Detecting Low-Contrast Texts on Web Pages at the Granular Character Level

User-friendly interfaces play a critical role in enhancing user experience within software development. As back-ground images, graphics, and intricately stylized text evolve in Web UI/UX design, testing for accessible visual appeal has become increasingly difficult. While evaluating text-background contrast may seem straightforward in simple cases, existing methods often fall short when faced with complex backgrounds, overlapping text, or exquisitely stylized text, thus motivating our research. To address these challenges, we propose a novel approach capable of detecting and visualizing low-contrast text at a granular character level through a pixel-level contrast-based clustering algorithm. Based on experimental findings from 10 real-world cases involving 1106 web elements, we've demonstrated the efficacy of our approach, achieving a high precision of 0.803 and a recall of 0.747.