Iterative, reference-conditioned image editing can leave behind grid-like and granular textures that researchers describe as digital ripple. A new paper posted on Hugging Face, titled Mi-Ripple: Restoring Images Degraded by Iterative AI Editing, proposes a workflow aimed at suppressing these artifacts while preserving image structure.
The method, described as a diagnosis-guided restoration workflow, separates periodic lattice artifacts from content-entangled granular texture. It then combines three components: selective spectral notching, structure-aware smoothing, and cleaned-reference regeneration.
According to the paper, this separation allows low-distortion filtering when artifacts are spectrally isolated, and visual reconstruction when filtering would otherwise erase legitimate detail. The authors report that across fourteen notch-only executions, whole-image residual standard deviation falls between 0.08 and 0.44 in CIELAB lightness units. In a paired regeneration example, reference cleaning reduced output debris density by 45%.
The work also investigates how resampling introduces periodic artifacts and why they can reappear after cleanup during subsequent editing. The authors frame Mi-Ripple as linking measurable artifact reduction to visibly cleaner generated images, rather than optimizing a spectral score alone. A project website is available at lab.miyang.cn/ripple.