This paper proposed a non-blind watermarking technique based on different wavelet decomposition levels for biometric image protection. In this technique, a biometric image is used as a watermark instead of a standard image, logo or random noise pattern type watermark. For watermark embedding, the original host image and the watermark biometric image are transformed into various levels of wavelet coefficients. The watermark biometric image is embedded into the host image by modifying the values of the wavelet coefficients of the host image using the values of wavelet coefficients of the watermark biometric image. Experimental results demonstrated that the proposed technique was able to withstand various watermarking attacks. The novelty of the proposed technique is that it is used to transform coefficients of the watermark biometric image instead of the Pseudo Noise sequences or any other feature extraction technique.