Robust object-based watermarking scheme via shape self-similarity segmentation

This study proposes a new object-based watermarking scheme based on efficient self-similarity segmentation in the spatial domain for solving the synchronizing detection problem. The lengths and orientations of two principle axes of image objects were used for image normalization and shape segmentation. Watermarks then can be embedded into object images by adjusting the pixel intensities on each segmented region. Experimental results reveal that the proposed method effectively resists geometrical attacks without using the original image. 2004 Elsevier B.V. All rights reserved.