In this paper, a robust digital image watermarking approach using adaptive quantization of wavelet packet coefficients was proposed. The original image is decomposed by discrete wavelet packet transform and the dominant wavelet coefficients are selected for watermark embedding from each sub-band except the lowest frequency one. Then, each watermark bit is adaptively embedded with different strength into the selected wavelet packet coefficients based on the odd or even value after quantization. Experimental results show that the proposed method is effective for copyright verification and robust to various common attacks such as Gaussian noise, JPEG compression, and without original image as references.