Abstract:

To preserve more original features after decolorization, this paper presents a new color-to-gray conversion algorithm based on multi-scale image fusion principles. The color image is decomposed into three monochrome channel images：R, G and B, the multi-scale image fusion model based on Gaussian-Lapacian pyramid is adopted to convert color image into gray image, the Gradient Domain Guided Image Filter（GGIF） is employed to eliminate the artifacts that may be produced by multi-scale fusion. The decolorization problem is transformed into a multi-scale fusion problem that preserves the image features of three monochromatic channels in a fusion image as far as possible. Experimental results show that the proposed algorithm is more sensitive to edge information compared with the existing state-of-the-art decolorization techniques, and can preserve the detailed features which are lost in the other methods in bright or dark areas.

Key words: color-to-gray conversion, image fusion, pyramid decomposition, Gradient Domain Guided Image Filter（GGIF）

摘要：

为了使彩色图像灰度化后能够保留更多的原始特征，提出了一种新的基于多尺度图像融合的灰度化算法。将彩色图像分解为R、G、B三个通道图像，采用基于高斯-拉普拉斯金字塔的多尺度图像融合模型进行灰度化，并引入梯度域导向图像滤波（Gradient Domain Guided Image Filter，GGIF）来消除多尺度融合可能产生的伪影。将灰度化问题转换为保持三个通道单色图像特征的多尺度融合问题。实验结果表明，与其他灰度化算法相比，所提算法对边缘信息敏感，并能够在较亮或较暗区域检测出更多的图像细节特征。

关键词: 灰度化, 图像融合, 金字塔分解, 梯度域导向图像滤波（GGIF）