Results ∥  Baseline ∥  Lambda ∥  Alphabet ∥  User Study 

  This paper presents an algorithm for creating Micrography QR codes, a novel machine-readable micrography generated by embedding a QR code into a micrography image. The unique structure of micrography renders existing methods designed for combining QR codes with natural or halftone images infeasible. We exploit the high-frequency nature of micrography to design a novel deformation model that skillfully warps the individual letters and adjusts their font weights to embed a QR code into a micrography. The whole process is supervised by a set of visual quality metrics tailored for micrography and a novel QR code quality measure to strike a balance between visual distortion and decoding correctness. A critical component to the success of our method is a QR code quality measure that predicts the chance of correctly decoding a deformed QR code. Particularly, the QR code quality measure is defined as a probabilistic model learned from decoding experiments on synthetic QR codes using popular decoders to capture various distortions due to image embedding. Experimental results show the strength of our method to generate high quality Micrography QR codes from a wide variety of inputs and enable diverse designs by embedding QR codes with varying sizes. We further conducted experiments and user studies to extensively evaluate each component in our method.