摘要： 提出一种基于自组织特征映射(SOFM)的灰度图像彩色化算法。将彩色样本图像和灰度图像均转换到lαβ颜色空间，在 通道构造像素邻域的多维特征向量，将彩色样本图像的像素亮度特征向量作为网络的输入进行训练，利用SOFM的自组织特性建立像素特征向量与彩色信息值的对应关系，应用训练好的网络实现灰度图像的彩色化。实验结果表明，该算法的彩色化效果较已有算法有较大改善。

关键词: 彩色化, 自组织特征映射, 金字塔分解

Abstract: A colorizing algorithm for grayscale images based on SOFM is proposed. The color image and grey image are first transformed into lαβ color space. In the l-channel, structure of the multi-dimensional pixel neighborhood feature vectors is constructed as the network input for training, and establishes the correlation between the pixel characteristics and color information using the self-organizing of SOFM. The trained network is used to colorize the gray image. Experimental results show that the algorithm results in the colorization are greatly improved than the previous algorithms.

Key words: colorization, self-organization feature map, pyramid decomposition

中图分类号: 

• TN911.73