An Example-based Color Transfer and Style Enhancement Through Locally Neighbor Embedding
Tao Liu, Daiguo Deng, Xuelian Wu, Kun Zeng · Advances in intelligent systems research/Advances in Intelligent Systems Research · 2013
Abstr act.Image color and tone adjustment are often used for style enhancement in image editing.In this paper we propose a new LLE-based image color transfer and tone enhancement method.Our target is to find the implicit color mapping rule between example image-pairs, then apply it to new image.Given a pair of images with the same size and theme but different color styles, we first divide them into lots of small patches and learn the implicit relationships among those patches using locally linear embedding(LLE).Then we impose the learned transfer pattern on input image, and take some extra processing techniques to make output image more natural.We apply our method for image color transfer andcolor images enhancement in these experiments, and the results show that our approach is effective.Keywor dsColor transfer•Tone enhancement•Locally linear embedding 1 Intr oduction Image styles are usually represented by image colors and tones.In the field of photography, a lot of works use large number of colors to generate special image styles and visual effects. in the field of art and design, people commonly do color editing and rendering to enhance the image display effects.But for those image editing workers, it is a tedious and boring task to match such large number of colors for images.If we are able to directly take good effect images as examples and learn colors and tones from them, it would be a meaningful work.Meanwhile, traditional algorithms do not perform very well in accurate local color transfer and little work is from a learning perspective.