Color filter arrays(CFA) based on complementary colors(CYMG) has been designed and used with the main advantage: higher spectral sensitivity and wider bandwidth than RGB CFA, especially in the low-light or short integration time environment. As for the interline-transfer, interlaced-readout complementary colors CFA pattern CCD, we propose a method of color restoration by the conversion of CYMG-YUV-RGB color space. Specifically, summing horizontally adjacent pixels from raw data can provide estimates of the luminance channel Y, while subtracting horizontally interlaced adjacent pixels can provide estimates of chrominance (color difference) channel U and V individually. A 2 × 2 pixels block of raw data is the smallest cell to figure out Y, U, V channels. Subsequently, transform YUV to RGB linearly according to the conversion formula between CYMG and RGB. At last, the raw data from CCD can be restored to RGB signals which is convenient for post-process, such as white balance. Additionally, we adopt an improved median filter to U and V channels to remove the edge zipper noise caused by interpolation, which can optimize the image quality.