Blue-Noise Sampling from Pixels to Graphs: Halftoning, Compressive Imaging, Structured Light, and Graph Signal Processing

Abstract

Blue-noise sampling patterns—originally developed for digital halftoning—exhibit two key properties: they suppress low-frequency artifacts while distributing samples as uniformly as possible in space. These characteristics make blue-noise patterns attractive far beyond print, including compressive imaging, 3D structured-light scanning, and modern graph signal processing. In this talk, I will first review blue-noise halftoning and its void-and-cluster construction, emphasizing how spatial dithering yields binary patterns whose Fourier spectra are dominated by high frequencies. I will then show how similar ideas inform coded-aperture and compressive spectral imaging, where blue-noise–like codes improve reconstruction quality and robustness. Next, I will connect these sampling principles to phase-measuring profilometry and related structured-light scanners, highlighting how pattern design impacts depth accuracy and calibration. Finally, I will present recent results on sampling of signals on graphs using blue-noise dithering, where spatially regular node subsets enable accurate reconstruction without explicit Laplacian eigendecompositions. Throughout the seminar, I will emphasize a common perspective: blue-noise sampling as a unifying design principle that links classical image halftoning to emerging applications in compressive sensing and graph signal processing, with opportunities for new collaborations in imaging, sensing, and data science.

Bio

Daniel L. Lau received his B.Sc. (with highest distinction) in Electrical Engineering from Purdue University in 1995 and his Ph.D. from the University of Delaware in 1999. He is the DataBeam Professor of Electrical and Computer Engineering and Director of Graduate Studies at the University of Kentucky, as well as an IEEE Fellow and Licensed Professional Engineer. Before joining academia, he worked as a DSP engineer at Aware Inc. and as an image and signal processing engineer at Lawrence Livermore National Laboratory. He also serves as Chief Technology Officer at Seikowave Inc., which develops 3D scanners for oil and gas pipeline inspection.