Abstract: It is possible to recolorize images by applying existing scribble based colorization algorithms for grayscale images, which omitting colors in original images. This paper proposes an optimized method to improve the existing image recolorization technologies. In comparison with the optimization-based colorization, the proposed method features in: （i） using K Nearest Neighbors（KNN） to preprocess images into stratified layers and learn from their content to produce a new simulated weight function; （ii） colorizing images in terms of the optimized layer-based weights and hence producing optimized colorizations. Extensive experimental results show that the proposed algorithm can solve the problem of color leakage at the boundary of the object, and obtain accurate color images. Compared with previous methods, the proposed algorithm is more robust to color blending in the input data.

Key words: optimization-based colorization, K Nearest Neighbors（KNN）, recolorization, layers information

摘要： 针对灰度图像彩色化技术应用于彩色图像二次着色时往往忽略掉原始图像所带的色彩信息的问题，提出了一种基于[KNN]图层区分的优化式着色算法。与现有的优化式着色方法相比，该方法一方面采用基于[KNN]的图像前背景区分算法获得图层区分的图像，生成新的权值函数；另一方面将图层区分结果引入优化式着色方法，并对图像着色。实验结果表明，算法能有效解决物体边界处发生颜色渗漏的问题，得到颜色分布精确的图像。在相同输入前提下，算法可以得到更好的着色结果。

关键词: 优化式着色, K最邻近结点算法（KNN）, 二次着色, 图层信息