Depixelizing Pixel Art: Johannes Kopf, Dani Lischinski.


We describe a novel algorithm for extracting a resolution-independent vector representation from pixel art images, which en-ables magnifying the results by an arbitrary amount without im-age degradation. Our algorithm resolves pixel-scale features in theinput and converts them into regions with smoothly varying shad-ing that are crisply separated by piecewise-smooth contour curves.In the original image, pixels are represented on a square pixel lat-tice, where diagonal neighbors are only connected through a singlepoint. This causes thin features to become visually disconnectedunder magnification by conventional means, and creates ambigui-ties in the connectedness and separation of diagonal neighbors. Thekey to our algorithm is in resolving these ambiguities. This enablesus to reshape the pixel cells so that neighboring pixels belongingto the same feature are connected through edges, thereby preserv-ing the feature connectivity under magnification. We reduce pixelaliasing artifacts and improve smoothness by fitting spline curvesto contours in the image and optimizing their control points.Keywords: pixel art, upscaling, vectorization