Progressive Color Transfer with Dense Semantic Correspondences

TL;DR

This paper introduces a novel color transfer algorithm that uses dense semantic correspondences and neural representations to achieve spatially variant, globally coherent color transfer between perceptually similar images, even extending to one-to-many mappings.

Contribution

It presents a new method that jointly optimizes semantic matching and color transfer using a coarse-to-fine approach, improving accuracy and extending to complex one-to-many scenarios.

Findings

Effective color transfer on diverse images

Handles one-to-many color transfer scenarios

Maintains spatial coherence and semantic consistency

Abstract

We propose a new algorithm for color transfer between images that have perceptually similar semantic structures. We aim to achieve a more accurate color transfer that leverages semantically-meaningful dense correspondence between images. To accomplish this, our algorithm uses neural representations for matching. Additionally, the color transfer should be spatially variant and globally coherent. Therefore, our algorithm optimizes a local linear model for color transfer satisfying both local and global constraints. Our proposed approach jointly optimizes matching and color transfer, adopting a coarse-to-fine strategy. The proposed method can be successfully extended from one-to-one to one-to-many color transfer. The latter further addresses the problem of mismatching elements of the input image. We validate our proposed method by testing it on a large variety of image content.