This academic article introduces a new method for image vectorization called Optimize & Reduce (O&R). The process of vectorization involves converting raster images into vector images. The proposed O&R method aims to provide a compact representation of input images by iteratively optimizing Bézier curve parameters and significantly reducing the number of shapes. The authors claim that their method is domain-agnostic and outperforms existing methods in both reconstruction and perceptual quality. They also highlight that their algorithm is faster than current optimization-based methods.
Publication date: 19 Dec 2023
Project Page: https://github.com/ajevnisek/optimize-and-reduce
Paper: https://arxiv.org/pdf/2312.11334
