快速黑白图像自动上色技术研究

The Research on Fast Algorithm for Automatic Grayscale Image Colorization

黑白图像自动上色是通过一种算法,让计算机能够自动地在两幅图像之间传递颜色信息,使目标图像(灰度图像)的颜色能够带有源图像(彩色图像)的特征。本文对目前常见的三种上色算法进行了研究与改进。基于灰度直方图的Hist算法上色速度非常快,作者提出基于邻域相关性的改进方法,使得该算法在保证上色速度的前提下提高了上色质量。金字塔算法是一种基于图像分层的算法,本文利用灰度直方图对图层上色进行了改进,从而大大提高金字塔算法的上色速度。快速上色算法是在经典算法Welsh算法的基础上,利用隔行扫描和邻域相关性进行了改进。

Grey image auto-colorizing is that the computer can automatically transfer color information between the two images through an algorithm and the target images （gray images）can be had color characteristics from source images （ color images）. Three algorithms have been studied and improved in this paper. We proposed improved method based on neighborhood relativity for histogram-based algorithm, this method improves quality of colorizing pictures without reducing the speed. Pyramid algorithm is a kind of algorithm basis on image layer, and this paper makes use of histogram improving layer colorizing, therefore increasing the speed of Pyramid algorithm. Fast algorithm based on Welsh algorithm is modified with interlaced scanning and neighborhood relativity.