Dithers of the Error-Diffusion Kind!
summary
A scoring algorithm blurs dithered and original images and compares them with structural similarity index measure to quantify quality across a Creative Commons Flickr faces dataset and line art; new kernels are generated by simulated annealing and evaluated quickly on small images. Random medium-score kernels produce dither patterns that fuse with subject matter or create vertical stripes, while some kernels beat Floyd‑Steinberg in a triangular pixels simulation; outputs include a web app, videos, room‑size installations and paintings.
AI-written summary · about
chapters (4)
related talks
31mPicasso, Geometry, Jupyter
16mPerceptual Evaluations: Evals for Aesthetics
08mDynamic Guidance for Decluttering Photographic Compositions
18mUsing Animation to Alleviate Overdraw in Multiclass Scatterplot Matrices
12mAlgorithmically generated flower beds!!
26mMaximizing Multimodal: Exploring the search frontier of text-to-image models
51mGoofing Off with Fun (Big) Data
33mDynamic Object Gaze Tracking with OpenCV