<p>Digital watermarking schemes have been used for copyright protection and tamper detection of various digital media, including digital images, videos, and audio. The problem with ordinary digital watermarking schemes for digital images is that they degrade visual quality of the images. There is a trade-off between the robustness of the digital watermark and the effect on the visual quality of the cover image. In contrast, zero-watermarking schemes that do not embed a watermark directly in the digital media have been proposed. Therefore, zero-watermarking schemes have no effect on the visual quality of the digital images. Nonetheless, they are as effective as non-zero-watermarking schemes. In this paper, we propose a zero-watermarking scheme for digital images. The proposed scheme is robust against various filtering and geometric attacks. The proposed scheme is based on calculating features from multiple sub-images of the original cover image. The selection of sub-images is based on a secret key. Therefore, feature extraction from the cover image is a function of the cover image itself and the secret key. Subsequently, the feature calculated from the cover image is merged with the encrypted version of the watermark image to obtain the final zero-watermark. Our experimental results show that the proposed zero-watermarking scheme is robust against JPEG compression, filtering, and geometric attacks. Since sub-image selection and the encryption of the original watermark are based on secret keys, only the authentic copyright owner of the digital image can extract the watermark from a watermarked image.</p>