A Novel Color Transfer Algorithm for Image Sequences

This paper presents a novel image sequence color transfer algorithm (ISCT). It is able to render an image sequence with color characteristics borrowed from three user-given target images. The input of this algorithm consists of a single input image (I1) and three target images (T1, T2, T3). The output of the algorithm is an image sequence {Si} with N images that embeds itself with color mood variations. A user selects necessary parameters, such as N and a color variation curve (CVC), through a user friendly interface we have developed. The algorithm completes the task in three steps. First, it executes a color transfer algorithm using the input image and three target images, producing three basis images (B1, B2, B3). Their color features are thus borrowed from the three target images (T1, T2, T3). Given the user-defined CVCs and three basis images generated in the first step, the ISCT algorithm then interpolates the mean and variance values for N images. Finally, referring to the input image and values interpolated in the second step, the algorithm automatically generates an image sequence. This is achieved in several seconds by applying the color transfer algorithm individually for every interpolated value. In addition, we have developed a user interface which makes it possible to view the rendered image sequence. Experimental results show that our algorithm swiftly produces an image sequence containing color variation with little user intervention. Given only three target images, the novel ISCT algorithm can demonstrate its ability to produce an image sequence with color mood variation as well as visually plausible effects. This algorithm is automatic, effective, and expeditious, and is appropriate for many applications.