Several people have built on top of this work, extending or improving parts of their proposed framework. Additional processing is needed to ensure spatial smoothness. They fit a template (optimal or arbitrary) over the image histogram, so they can shift hues accordingly to harmonize colors or composites from several sources. use hue histograms and harmonic templates defined as sectors of hue-saturation in HSV color space, to model and manipulate color relationships. Many existing works have applied different concepts from traditional color theory for artists to improve the color composition of digital images. Our algorithm is a key contribution which enables our approach and many other Very high resolution images ( ≥ 100 megapixels)-20x faster than the state-of-the-art [Aksoy Our algorithm’s performance is extremely efficient even for The user can edit the palette and obtain new decompositions instantaneously. Which captures color as well as spatial relationshipsĪfter an initial palette is extracted (given an RMSE reconstruction threshold), We consider the geometry of 5D RGBXY-space, Our approach is inspired by the geometric palette extraction technique of Tan et al. We propose a new, extremely efficient yet simple and robust algorithm to do so. Images are decomposed into a palette and associated per-pixel compositing or mixing parameters. In this paper we focus on monochrome, complementary, single split, triad, double split, square and analogous.Īt the core of our and other recent approaches [Chang et al., 2015 Tan , although some of them have a secondary degree of freedom α 2 that enforces symmetry. Each template can be modeled by a single global rotation α 1 Although we don’t use it in our results, our approach can also describe hybrid templates like L type. These templates are popular among creatives, but they are also in agreement with the definitions of similarity and ambiguity by Moon and Spencer. Note that our templates derived from separate Y type into single split and triad, and the same for X type. We found this distinction helps handling templates like analogous properly. We use two different types of axes: the dashed ones attract colors towards them the solid ones define sectors between them, so colors inside remain in the same position, but colors outside are attracted towards them.

Comparison between the sector-based hue harmonic templates from (shaded in grey), and our new axis-based ones. Our proposed framework can be used by other palette-based systems and workflows,Įither for palette improvement or image editing. We also propose new color operations using this axes-based representation. (appropriate for hue histograms) we derive our harmonic templates from classical color theory (see Figures 2 and 14). Instead of using the sector-based templates from Matsuda [Tokumaru

We demonstrate applications to color harmonization and color transfer. This is useful for various color-aware tasks. This last point enables the simplification of harmonic templates and other relationships into a set of a few 3D axes that capture color structure in a meaningful and compact way. Color palettes also simplify the modelling and formulation of relationships between colors.

Color palette from image 1.5