In the 21st century, the use of internet and multimedia applications is growing day by day. The demand of online trade and digital content, in the form of audio, video, text, images is increasing. But the problem arises that the digital data can easily be manipulated and duplicated. This increased the demand of securing the digital data for copyright protection, preventing unauthorized distribution, and ownership verification. Digital image watermarking is a powerful tool to secure digital data. An information called ownership identifier is hide in the original data with a suitable algorithm so as not to affect the quality of the original data. Watermarking techniques mainly aims in maintaining the robustness and imperceptibility of the image. Watermark can be embedded in frequency domain or spatial domain. Frequency domain is the mainly used technique due to its robustness against different types of attacks. Watermark can be extracted using blind and non-blind technique. In this paper, 3-level DWT –SVD of the host image is done by Haar wavelet and the technique is used to embed watermark in the host image. Further a median filter is used to improve PSNR at a low value of alpha. Watermark is extracted using blind technique. The experimental results conclude that the extracted watermark with the proposed algorithm is invisible and quality of watermarked image and recovered image is good.