Abstract: Local color similarity is the basis of image colorization, where choosing the suitable color space is a key problem. According to the standard whether luminance and chrominance could be separated from each other, common color spaces were divided into two categories, i. e. color-component space and color-attribute space. Smoothness of color distribution differs in different color spaces. The applicability of different color spaces in colorization was evaluated. Theoretic and experimental results show that color-attribute space is more appropriate to use in colorization. Furthermore some conclusions on color space selection are drawn by analyzing the essence of colorization.

Key words: Colorization, Color space, Application analysis