We propose a novel localized geometric-distortion resilient digital watermarking scheme to embed two invisible messages to images. Our scheme utilizes local circular region- (LCR) based and block DCT-based techniques to hide two watermarks in two kinds of regions and extract these two embedding sequences from their individual regions. Specifically, we use the histogram and mean statistically independent of pixel positions to embed one watermark in LCRs, whose centers are SIFT feature points that are robust against various attacks. We also use Watson's DCT-based visual model to embed the other watermark in several rich textured regions not covered by any embedding LCR. Our experimental results demonstrate the proposed system achieves better performance than several peer systems in the literature.