Paper widgets: Visually aesthetic “smarts” for document images

Traditional 1D and 2D barcodes provide high data density, but they are visually jarring and require isolated white margins for placement. In this work, we introduce new machine-readable "smarts" for paper documents, called paper widgets. Like barcodes, paper widgets contain digital data which can be read upon scanning and decoding. However, unlike barcodes, they have very small footprint (fraction of sq.cm.), and carry a human-readable component that provides visual meaning. Furthermore, paper widgets can be placed and recovered in a distributed fashion from any position on a document image. For example, they can be positioned right beside contents of interest, and need not be confined to isolated white margins. We describe how all these widget properties can be simultaneously achieved using simple operations that are robust to print-scan distortions. In particular, we highlight a novel constrained coding technique that helps combat print-scan intersymbol interference (ISI), and a Gabor-filtering based extraction that accurately identifies widget regions from any position on a scanned document image. Experimental evaluations reveal that paper widgets can be recovered from print-scan distortions with near-100% accuracy.