Block based copy move image forgery detection methods usually have problems such as high dimensional feature vector, high computational complexity and low detection rate.So, this paper proposed a new method for image copy move forgery detection based on color invariants, which extracts density characteristics from each channel of opponent color space converted from RGB. The k d tree is constructed to speed up approximate nearest neighbors matching for tampering detection. The experiment results show that the proposed local color invariants feature with low dimension and simple computation can represent the image block feature effectively. When compared with typical block based methods, this method has some advantages, such as lower computational complexity, more accurate and robust to post processing for forgery regions such as rotation and scaling.