Color Transfer Algorithm

With the high development of the computer hardware and multimedia technologies, image processing has played a big role in people's daily life. Moreover, the technology of color image processing influences people directly. Although it has become the hot topic in areas of image processing and computer vision, most of algorithms about color images are only the extension of algorithms about grayscale images. They lack some special aims at color image.Based on all above factors, this paper dose some research about Color Transfer which is a new algorithm created in 2001. The input of Color Transfer is one source color image and another target image. The output is an image owning both target image's outline and source image's color sense. And in recent years, researchers abroad or natively have proposed many advanced algorithms of color transfer which aim at promoting transferring result or being applied in a new area. All these can prove that the algorithm can be applied to produce a new picture, remove fog and make special effect.This paper firstly introduces the process of original color transfer algorithm, gives some experimental results in three special aspects, analyses its some drawbacks and summaries two classes of improving algorithms in recent years: the improvement of transferring method and the improvement of color segmentation. Then, after discussing some color spaces and some improving algorithms about transferring process, this paper propose an improved PCA algorithm to transfer color between images and try to apply ICA to complete color transfer. About the improvement of color segmentation, this paper analyses some exist manual and auto segmentation methods and propose to apply FCM to segment the source and target images to transfer color with corresponding regions.Finally, this paper transfers the color of segmented face and hair regions of real-time video with advances methods and a segmented make-up face picture. It can avoid the error color of background in original algorithm.For the most algorithms in our paper, we do a lot of experiments under OpenCV1.0+VC6.0 or Matlab. The experiment results of special aspects can prove the practicability of color transfer algorithm. Meanwhile, the contrast between the results of original algorithm and improved algorithms can prove the validity of our algorithms.