Recomposed realities: animating still images via patch clustering and randomness

TL;DR

This paper introduces a patch-based method that animates still images by clustering patches from datasets and reconstructing images through random sampling, enabling lively reinterpretations of static images.

Contribution

It proposes a novel patch clustering and sampling technique for animating still images, emphasizing reinterpretation over replication.

Findings

Effective in animating diverse still images

Preserves local structures while allowing conceptual differences

Enables reinterpretation through patch clustering

Abstract

We present a patch-based image reconstruction and animation method that uses existing image data to bring still images to life through motion. Image patches from curated datasets are grouped using k-means clustering and a new target image is reconstructed by matching and randomly sampling from these clusters. This approach emphasizes reinterpretation over replication, allowing the source and target domains to differ conceptually while sharing local structures.