| 3-D models can be represented as meshes and point clouds, whereas 3-D motion data is defined by time varying 3-D points for every human joint or as a matrix which is multivariate and multi-attribute. Representation, data processing operations and dimensionality of these data sets makes it difficult to reuse of existing content protection (copyright protection and tamper detection-correction) watermarking and finger printing based methods related to images, video, text and audio. Therefore, this dissertation addresses the issues related to the design of schemes for 3-D content protection. Copyright protection techniques require robust watermark to withstand data processing operations, whereas watermarking and finger printing schemes aim to accurately identify tampering due to data processing operations.; To handle the case of robust watermarks and build generic schemes for different representation of 3D contents, clustering based spatial robust blind watermarking mechanisms have been proposed. In order to encode watermark related bits for a given set of n points with n! possible orders, we need to find an order. To find such an order, ordered groups of clusters of 3-D points are identified on the basis of time for motion data and proximity in 3-D space for 3-D models. Inside the clusters, 3-D based scalar quantities order the points locally and bits can be encoded or decoded using proposed extensions to 3-D quantization index modulations. The schemes are analyzed for robustness against uniform affine transforms (scaling, rotation and translation), cropping and reordering. Comparatively, the schemes are less robust against randomized noise addition and simplification attacks. These can be improved by using benchmarking in case of motion data, and encoding more bits per point for 3-D models. The encoding schemes can be customized to achieve high hiding capacity, imperceptible watermarks and security.; For tamper detection of motion data, large imperceptible fragile watermarks are encoded inside time based clusters, using 3-point and 1-point quantization index modulation based bit encoding. The watermarks can accurately detect and classify attacks such as order reversal, uniform affine transforms and noise addition. This scheme is improved by encoding watermarks in the localized domain, and benchmarking the bit encoding scheme. Watermarks are not lost during semantic invariant operations (uniform affine transform and noise additions), meanwhile detecting attacks such as noise addition, affine transforms and order reversal. Further, tamper detection probable correction methodology has been proposed that combines fragile watermarking and finger-printing. These schemes identify shuffling attacks on rows, columns or their elements and correct them. Random noise addition attacks can be detected using fragile watermarks and corrected using interpolation.; For tamper detection in 3-D models, the skeleton of the 3-D model is authenticated against cropping and noise addition attack, using a customized version of the clustering based approach for 3-D models. Watermarks are not lost during progressive compression-decompression and uniform affine transform, when the skeleton stays unaltered. |
