Color has always been a major contribution to the mood of an art piece and even more when dealing with abstract art. "Omoiiro" is a collection of new and old (but finely tuned) techniques that alleviate the difficult process of choosing colors by automatically generating color palettes from various sources. It could start from a single hue value (this web site is using colors generated by such technique) or the algorithms can extract aesthetically pleasant combinations of colors from existing images (mostly photos). "Omoiiro" ranges from simple methods (so very fast, for interactive applications) to complex methods that leverage our understanding of the human perception of colors, exploring the distribution of the colors in various spaces before selecting the best candidates. The fine tuning of all the methods reflects the aesthetic sense of this project’s researcher, Alexis Andre.

For example Figure 1 shows palettes based around one hue. The idea is to create colors "around" that particular tone and to provide secondary colors for accents.

Fig.1 Color palettes from single hue values.

Figure 2 shows the most basic color extraction method by randomly sampling pixels in the original image. This results in simple yet nice palette that eventually will cover the ranges of colors in the image.

Fig.2 Randomly selected colors (right) from pictures (left).

Finally, when thinking about color from a human perception point of view, palettes that contains only discernible colors can be created, without the risk of losing some accent colors in the picture since the methods is looking in a specific color space unrelated to the picture. Figure 3 shows one of the most advanced algorithms in the "Omoiiro" series: clustering in Lab space with adaptable number of clusters that result in nice palettes that reflect the mood of the original image.

Fig.3 Clustering in Lab color space: the pictures on the left are the source of the palettes generated on the right.

(All pictures taken by Alexis Andre in Morocco/Spain)

Go to top