Creating a Color Palette from an Image
Summary
This article documents Spectrimage's approach to extracting a five-color palette from an image, detailing iterative improvements from RGB median-cut to OKLCH-based clustering with K-means++, and a multi-pass refinement including phantom guard and mass-based slot allocation to produce human-friendly palettes. It provides practical thresholds, visualization of results, and notes on user-adjustable swatches in the Spectrimage app.