This paper presents a robust color image hiding method based on YCbCr color system in discrete cosine transform (DCT) domain, which can hide a secret color image behind a public color cover image and is compatible with the international image compression standard of JPEG. To overcome the grave distortion problem in the restored secret image, this paper proposes a new embedding scheme consisting of reasonable partition of a pixel value and sign embedding. Moreover, based on human visual system (HVS) and fuzzy theory, this paper also presents a fuzzy classification method for DCT sub-blocks to realize the adaptive selection of embedding strength. The experimental results show that the maximum distortion error in pixel value for the extracted secret image is ± 1. And the color cover image can provide good quality after embedding large amount of data. This work is partially supported by National Science Foundation (Grant No. A0324676)