Palette-Based Content-Aware Image Recoloring

Palette-based image recoloring is a popular image editing technique in recent years. It allows users to perform global color edits to an image by manipulating a small set of representative colors. Many approaches have been proposed for palette extraction and palette-based image recoloring. However, existing methods primarily leverage low-level visual information to extract color palettes, so that different objects with similar colors will share the same palette colors. It is impossible to individually recolor one of multiple objects with similar colors, as altering a specific palette color may cause unwanted color changes to many non-interesting objects. To address this issue, in this paper, we present a novel, palette-based content-aware image recoloring approach. Different from previous methods, we extract the color palette of an image in a high-dimensional space that combines low-level visual features and high-level semantic features, allowing generating separate palette colors for different objects with similar colors. This enables users to perform targeted local editing, i.e., distinguish and recolor objects with similar colors separately, without producing unexpected global color changes. Extensive experiments demonstrate the flexibility, local control, and effectiveness of our method.