Authors:
Ali Javid
and
David Mould
Affiliation:
School of Computer Science, Carleton University, Ottawa, Canada
Keyword(s):
Animation, Image Processing, Palette Cycling, Pixel Art.
Abstract:
We propose an automated method for palette cycling, a technique for animation storage and playback. A palette cycling animation uses a fixed map of indices over the entire animation; each frame, the color lookup table accesses by the indices changes. Historically, palette cycling animations were created manually. Here, we present a method that automatically creates a set of palettes and an index map from an input video. We use optimization, alternating between phases of choosing per-frame palettes and determining an index map given a fixed set of palettes. Our method is highly effective for scenarios such as time-lapse video, where the lighting changes dramatically but there is little overall motion. We show that it can also produce plausible outcomes for videos containing large-scale motion and moving backgrounds; palette rotations with these features are especially difficult to craft by hand. We demonstrate results over a variety of input videos with different levels of complexity,
motion, and subject matter.
(More)