This paper proposes a watermark insertion and detection scheme using Krawtchouk moments. These moments are selected for image watermarking because image reconstruction with these moments is better than other orthogonal moments like Legendre, Zernike and Tchebichef. The watermark signal is obtained by modifying some of the Krawtchouk moments of the original image. Further, the strength of the watermark is adjusted based on a threshold value(T) which is obtained by maximum root mean square error(rms value) between Krawtchouk moment invariants of the original image and Krawtchouk moment invariants of original image subjected to possible distortions like rotation, scale and translation. Krawtchouk moment invariants are used since they are invariant to image rotation, scale and translation. In order to detect the added watermark, we calculate the rms value between Krawtchouk moment invariants of test image and Krawtchouk moment invariants of the watermarked image. The watermark is detected if the above rms value is less than a threshold (T). It is noted from the simulation results that the proposed approach is robust with respect to geometric distortions like scale, translation and rotation.