Abstract

In this talk, I propose a simple interactive technique for adjusting the global color balance of an image by finding and replacing color gradients. The aim is to make it easier to control interactive color correction, a common user task in digital photography and video processing. For instance, the technique can assist in compensating for scene illumination, selectively emphasizing scene details, and applying a color scheme to impart a particular mood or theme. It may also prove useful in calibrating the mapping between the disparate color gamuts of different capture and display devices. In information visualization, color adjustment plays an important role in post-processing, where color distinctions are used to direct the viewers attention to the relevant structure of the data. The technique enables the user to improve the color composition of an image by choosing a harmonious combination of colors and adjusting their contrasts to highlight the key elements of the picture.

The user designates a set of source color gradients and a matching set of target color gradients to guide the algorithm in performing a continuous transformation of the image colors. Compared with previous work, this approach benefits from operating on color spans rather than individual colors. While a mapping between colors suffices only to define a translation in color space, a mapping between color gradients defines a more general affine transformation. The method works by applying a feature-based warping technique to calculate a nonlinear volumetric deformation of the color space. In effect, the resulting image can be seen as a composite of the original image and its color gradient transformations, with the mask of each layer determined by the region of influence of its color gradient.

The results of this technique will be discussed with a view towards the aesthetic considerations of color and contrast in graphic design. I will also briefly introduce other work in interactive image enhancement developed at the Computer Laboratory of the University of Cambridge, including contrast brushes for local contrast enhancement and histogram warping for global contrast enhancement.