NIFTY: a Non-Local Image Flow Matching for Texture Synthesis

ABSTRACT This paper addresses the problem of exemplar-based texture synthesis. Experimental results demonstrate the effectiveness of the proposed approach compared to representative methods from the literature. Index T erms-- Generative model, Image synthesis, Texture synthesis, Flow Matching 1. INTRODUCTION AND RELA TED WORK Image generative modeling has been a very active domain over the past decade, driven by a combination of theoretical and technical advances. This progress has led to the development of diverse generative modeling frameworks, many of which rely on the training of deep neural networks. Diffusion models Generative diffusion models (DMs) [1, 2, 3, 4, 5, 6, 7] have recently attracted significant attention for their ability to capture complex data distributions and generate high-quality samples, while benefiting from the stable training provided by conditional U-Net architectures.