针对传统颜色传递算法只进行全局颜色传递, 未考虑图像局部特征, 出现颜色误传递的问题, 提出了一种基于Graph-Based图像分割与区域匹配的灰度图像彩色化算法。利用Graph-Based图像分割将目标图像和参考图像分割为若干个亮度和纹理特征一致的子区域。根据本文提出的思想, 认为“具有相似亮度和纹理的子区域具有相似的颜色分布”, 对目标图像和参考图像的子区域建立匹配关系, 利用亮度特征和LBP局部纹理特征对子区域进行匹配。最后, 在YCbCr颜色空间中对相匹配子区域利用亮度和纹理特征进行颜色传递。实验结果证明, 在该算法中其参考图像无需和目标图像配准, 只需特征相近, 就可以对目标图像实现全自动的彩色化技术并有效克服了颜色误传递的现象,彩色化结果图像在色彩分布和主观视觉上都有明显的提高。

Aiming at the problem of the color mistransmission of the traditional color transfer algorithm which only carries out the global color transfer and doesn’t consider the local features of the image. A gray image colorization algorithm based on Graph-Based image segmentation and region matching is proposed. The target image and the reference image are segmented into a plurality of sub-regions whose luminance and texture features are consistent by using Graph-Based image segmentation. According to one idea presented in this paper, it is considered that "sub-regions with similar brightness and texture have similar color distributions", a matching relationship is established between the target image and the sub-region of the reference image, and the sub-region is matched by using the luminance feature and the LBP local texture feature. Finally, the color matching is performed on the matching sub-regions using luminance and texture features in the YCbCr color space. The experimental results show that the reference image does not need to be registered with the target image in the algorithm, and only the features are similar, the automatic colorization technology can be realized for the target image and effectively overcome the phenomenon of color mistransmission. The colorized result image has a significant improvement in color distribution and subjective visuals.