Research On Color Transfer Between Images

The results of image Color transfer is mainly reflected in : (1) Effect. Whether the color is natural and the picture is vivid is two most important measurements; (2) Efficiency. Both artificial participation and algorithm complexity are two aspects affecting efficiency. Using existing algorithms to process image, the junction region is unnatural, which requires better algorithm. Although some algorithms can bring out partly automation, in processing, user interaction is still inevitable. Comparing with artificial participation, algorithm complexity can obviously accelerate the processing speed. In the paper, research begins with these two areas.The algorithm in this paper is based on territory value segmentation of the junction region. The algorithm makes use of edge detection to distill the region of the edge, and conducts domain segmentation, then comminutes the pixels of regional at the junction each region strictly, transfer the similar color to the same region. Finally, the target image mean-variance is normalizedAlthough user interaction is, to some extent, able to improve the accuracy of color transferring, the efficiency of their implementation is too low, and has certain subjectivity. Automatic color transfer effectively solves this issue. Carve up Roseau and subtractive clustering is applied in this paper. Completely automatic algorithm implementation is realized. In clustering, pixels are selected according to their brightness, and those with larger data density are cluster to sample block. First of colors in the same block are transferred first, others are transferred later. This algorithm can be used to transfer color not only between two colorful images, but also between a grey image and a colorful image.Finally, this paper designs an image color transfer software system. The system is divided into user interaction color transfer interface and automatic color transfer interface. And this system is a very user-friendly one.