Visual Detection — All Articles & Tutorials

Visual detection is not just about comparing two images pixel by pixel. Depending on the type of defect targeted, algorithms and thresholds differ: a subtle color drift (mis-resolved token, sRGB / P3 conversion) calls for perceptual comparison; a layout shift (broken flexbox, duplicated margin) is better seen through structural analysis; a typography issue (fallback font loaded instead of the primary one) demands particular attention to sub-pixel rendering. And everything changes again in responsive or dark mode, where every breakpoint multiplies the surface to cover.

This page gathers articles that break down the types of detectable visual regressions and how to distinguish them: false positives caused by font smoothing, real contrast regressions that go unnoticed by the naked eye, layout breakages on narrow viewports, partial dark mode failures where some components forget to adapt. We also cover the choice of comparison algorithms (pixel matching, SSIM, perceptual) with their respective strengths and blind spots. Delta-QA relies on a combination of these approaches to minimize noise while staying strict on real regressions; these articles explain the principles to help you understand what a visual testing tool can — and cannot — reliably detect.