Parag Agarwal

Robust blind watermarking mechanism for point sampled geometry

Watermarking schemes for copyright protection of point cloud representation of 3D models operate only on the geometric data, and are also applicable to mesh based representations of 3D models, defined using geometry and topological information. For building such generic copyright schemes for 3D models, this paper presents a robust spatial blind watermarking mechanism for 3D point sampled geometry. To find the order in which points are to be encoded/decoded, a clustering approach is proposed. The points are divided into clusters, and ordering is achieved using inter-cluster and intra-cluster ordering. Inter-cluster ordering achieves local ordering of points, whereas intra-cluster ordering does it globally. Once ordered, a sequence of clusters is chosen based on nearest neighbor heuristic. An extension of quantization index of bit encoding scheme is proposed, and used to encode and decode inside the clusters. The encoding mechanism makes the technique robust against uniform affine transformations (rotation, scaling, and transformation), reordering attack and topology altering (e.g. retriangulation) attack when applied to 3D meshes as well. Replication of watermark provides robustness against localized noise addition, cropping, simplification and global noise addition attacks. Security of the scheme is analyzed, and the time complexity is estimated as O (n log n), where n is the number of 3D points. Theoretical bounds on hiding capacity are estimated, and experiments show that a high hiding capacity is high, with embedding rate greater than 3 bits/point. The bit encoding method reduces the distortions and makes the watermark imperceptible, indicated by a signal to noise ratio greater than 100 dB.

Robust Blind Watermarking of Point-Sampled Geometry

Digital watermarking for copyright protection of 3-D meshes cannot be directly applied to point clouds, since we need to derive consistent connectivity information, which might change due to attacks, such as noise addition and cropping. Schemes for point clouds operate only on the geometric data and, hence, are generic and applicable to mesh-based representations of 3-D models. For building generic copyright schemes for 3-D models, this paper presents a robust blind watermarking mechanism for 3-D point-sampled geometry. The basic idea is to find a cluster tree from clusters of 3-D points. Using the cluster tree, watermarks can be embedded and extracted by deriving an order among points at global (intracluster) and local levels (intercluster). The multiple bit watermarks are encoded/decoded inside each cluster based on an extension of the cluster structure-based 3-D quantization index modulation. The encoding mechanism makes the technique robust against uniform affine transformations (rotation, scaling, and transformation), reordering, cropping, simplification, and noise addition attacks. The technique when applied to 3-D meshes also achieves robustness against retriangulation and progressive compression techniques. Customization of the bit-encoding scheme achieves high hiding capacity with embedding rates that are equal to 4 b/point, while maintaining the imperceptibility of the watermark with low distortions. The estimated time complexity is O (n logn), where n is the number of 3-D points.

Robust blind watermarking mechanism for motion data streams

The commercial reuse of 3D motion capture (Mocap) data in animation and life sciences raises issues with respect to its copyright. In order to improvise content protection of Mocap data, we devise a substitutive blind watermarking technique. This technique visualizes 3D Mocap data as a series of non-intersecting cluster of triangles. Bits are encoded inside the triangles by using an extended substitutive bit encoder in spatial domain. The encoding supports watermark imperceptibility and develops robustness against affine transforms (rotation, translation, scaling), noise addition, reordering and sample loss attacks. Security of the scheme can be enhanced by adding secret embedding distances between clusters, which are based on a secret key used for watermarking purpose.

Blind robust watermarking of 3d motion data

The article addresses the problem of copyright protection for 3D motion-captured data by designing a robust blind watermarking mechanism. The mechanism segments motion capture data and identifies clusters of 3D points per segment. A watermark can be embedded and extracted within these clusters by using a proposed extension of 3D quantization index modulation. The watermarking scheme is blind in nature and the encoded watermarks are shown to be imperceptible, and secure. The resulting hiding capacity has bounds based on cluster size. The watermarks are shown to be robust against attacks such as uniform affine transformations (scaling, rotation, and translation), cropping, reordering, and noise addition. The time complexity for watermark embedding and extraction is estimated as O(n log n) and O(n2 log n), respectively.

Tamper proofing mechanisms for motion capture data

Repositories of motion captured (Mocap) data can be reused for human motion analysis in physical medicine, biomechanics, and animation related entertainment industry. Mocap data expressed as a matrix can be subject to tampering from shuffling of its elements or change in element values due to motion editing operations. Tampering of the archival system intentionally or due to machine/human errors, may result in loss of research, money and effort. In order to detect and correct errors induced due to tampering; this paper proposes a tamper proofing methodology that combines hash function and watermarking based methods. These patterns (fingerprints) resulting from hash functions help in error detection by identifying the type of attack such as row shuffling, column shuffling, row element shuffling, column element shuffling and their combinations. Random attacks that change data element values are detected by change in watermarks embedded in data elements. Finger prints help in solving the attacks reversal such as column shuffling and element shuffling, whereas watermarking helps in reversing attacks such as column element or row element shuffling. As compared to other attacks, random attacks cannot be reversed, and can be improved using interpolation. Analysis shows that the proposed method uses O(n) space to detect and correct errors, and the time complexity for correction varying from o(n log n) to O(n!).

Progressive compression invariant semi-fragile watermarks for 3d meshes

In this paper, we propose a scheme to enhance the robustness of authentication based watermarking algorithms for 3D models (3D meshes) against compression and decompression. We target different compression techniques such as Compressed Progressive Meshes (CPM) [9], Progressive Forest Split (PFS) compression [11], and Progressive Mesh (PM) [7]. The idea behind the algorithm is to compute those set of vertices that are never used for compression. We have three sets of such vertices that are never used for compression by these techniques. The first set of vertices includes the boundary vertices that are critical for deciding the shape of the 3D model. The second set includes the neighboring vertices to the split vertex. The third set of vertices comprise of edges that do not form simple triangle. After applying the compression algorithms for 3D models, we apply the watermarking algorithm that uses geometric information only. We then watermark the vertices that are obtained from the above three sets. These watermarks are shown to be preserved even after repeated decompression and compression of 3D models. Due to the generic nature of the algorithm, we can apply the above idea to any watermarking algorithm used for other applications such as copyright protection as well.

Minimizing probable collision pairs searched in interactive animation authoring

Animation authoring involves an author’s interaction with a scene, resulting in varying scene complexity for a given animation sequence. In such a varying environment, detection and prediction of collision in minimal time and with high accuracy is a challenge. This paper proposes using the bounding volume-based space subdivision mechanism to reduce search space for an object pair collision search. This data structure is enhanced using a direction-based spatial hash table, which predicts collision between static and dynamic objects. These techniques are shown to work in conjunction with existing search space reduction methods. The event of collision is accurately detected using known methods, such as kinetic data structures. Simulation results show that for a scene with 10000 objects with varying dynamic objects (10–90%), the method finds probable collision-pairs with 95–99% accuracy.

Reliable Transmission of Audio Streams in Lossy Channels Using Application Level Data Hiding

The paper improves the reliability of audio streams in a lossy channel. The mechanism groups audio data samples into source and carrier sets. The carrier set carry the information about the source set which is encoded using data hiding methodology - quantization index modulation. At the receiver side, a missing source data sample can be reconstructed using the carrier set and the remaining source set. Based on reliability constraints a hybrid design combining interleaving and data hiding is presented. Experiments show an improved reliability as compared to forward error correction and interleaving.

Tamper proofing 3d motion data streams

This paper presents a fragile watermarking technique to tamper proof (Mocap) motion capture data. The technique visualizes 3D Mocap data as a series of clusters of points. Watermarks are embedded using clusters of points, where a bit is encoded in each cluster. The four point encoding mechanism uses a combination of one point encoding and three point encoding schemes. Using these schemes it is possible to distinguish between affine transformations, noise addition and reverse ordering attacks. The bits are encoded and decoded in this scheme using an extension of quantization index modulation. It has been shown that distortions are reduced to achieve imperceptibility of the watermark. The bit encoding schemes give the flexibility to achieve better accuracy in tamper detection. In addition, the paper suggests a probabilistic model, which is a function of the watermark size. Using this model, it has been proved that larger watermark sizes achieve higher accuracy in tamper detection.