摘要
提出了图像局部彩色化的全局扩展算法,在给样本块上色之后,将图像纹理特征和亮度统计值作为图像特征向量分量,用于建立与样本块的对应关系;利用图像邻域信息的偏度和峰度,根据样本块对整幅图像聚类;利用查找每个像素点邻域像素所属聚类域的方法确定此像素点所属聚类域,在对应样本块中选取最佳匹配像素,实现目标图像全局彩色化,使彩色化结果图像在不同颜色边界的过渡更加自然、平滑、流畅。
The global expansion algorithm of local image colorization is presented in this paper. We transfer colors to swatches,and take texture characteristic and luminance of image as a feature vector component to construct corresponding relation of target images and swatches. Using the skewness and durtosis of neighborhood information of image,according to swatches,we do whole image clustering. And we select the best match pixel in the corresponding swatch, to achieve the overall image colorization. Colorized images by this method are more natural on the borders of different colors.
出处
《机械工程与自动化》
2009年第3期84-85,88,共3页
Mechanical Engineering & Automation
关键词
图像局部彩色化
全局扩展
聚类
local image colorization
global expansion
skewness
kurtosis
clustering
