Abstract Scientometrics has many citation‐based measurements for characterizing diversity, but most of these measurements depend on human‐designed categories and the granularity of discipline classifications sometimes does not allow in‐depth analysis. As such, the current paper proposes a new measurement for quantifying journals' diversity by utilizing the abstracts of scientific publications in journals, namely topic diversity (TD). Specifically, we apply a topic detection method to extract fine‐grained topics, rather than disciplines, in journals and adapt certain diversity indicators to calculate TD. Since TD only needs as inputs abstracts of publications rather than citing relationships between publications, this measurement has the potential to be widely used in scientometrics.

A new model, named as deep dynamic poisson factorization model, is proposed in this paper for analyzing sequential count vectors. The model based on the Poisson Factor Analysis method captures dependence among time steps by neural networks, representing the implicit distributions. Local complicated relationship is obtained from local implicit distribution, and deep latent structure is exploited to get the long-time dependence. Variational inference on latent variables and gradient descent based on the loss functions derived from variational distribution is performed in our inference. Synthetic datasets and real-world datasets are applied to the proposed model and our results show good predicting and fitting performance with interpretable latent structure.

In this paper, we propose to use contexts of superpixels as a prior to improve semantic segmentation by the CRF framework. A graphical model is constructed on over-segmented images. Our main contribution is to take the concept of “superpixel embedding” into consideration, which is formalized as a potential item for optimizing the energy of the whole graph. We also introduce two ways of calculating this embedding potential. Experiments on several popular datasets, e.g., MRSC-21 and PASCAL VOC, illustrate that our approach enhances the performance of a previously proposed segmentation model without embedding. The accuracy results are comparable to some fully supervised methods.

Motion capture (mocap) data normally used to produce a three-dimensional (3D) mesh animation is exploited for watermarking. Copyright protection of the 3D mesh animation obtained is not guaranteed since identical joints in the mocap data are not distinguished easily. A novel robust blind watermarking method is proposed for the copyright protection of 3D mesh animations in the paper. At the beginning, the shape diameter function (SDF) method segments one frame of a 3D animation into several meaningful parts. A watermark is then embedded into these parts by adjusting the geodesic distances. Vertex trajectories are further exploited to modify the geodesic distribution of residual frames to be the same as that of the watermarked frame. Our experimental result demonstrates that the proposed method is able to resist the noise addition, similarity transformation, quantization, cropping, simplification, smoothing, and resampling attacks.

A novel feature region selection method for robust digital image watermarking is proposed in this paper. This method aims to select a nonoverlapping feature region set, which has the greatest robustness against various attacks and can preserve image quality as much as possible after watermarked. It first performs a simulated attacking procedure using some predefined attacks to evaluate the robustness of every candidate feature region. According to the evaluation results, it then adopts a track-with-pruning procedure to search a minimal primary feature set which can resist the most predefined attacks. In order to enhance its resistance to undefined attacks under the constraint of preserving image quality, the primary feature set is then extended by adding into some auxiliary feature regions. This work is formulated as a multidimensional knapsack problem and solved by a genetic algorithm based approach. The experimental results for StirMark attacks on some benchmark images support our expectation that the primary feature set can resist all the predefined attacks and its extension can enhance the robustness against undefined attacks. Comparing with some well-known feature-based methods, the proposed method exhibits better performance in robust digital watermarking.

Pose editing, changing the poses of three-dimensional (3D) mesh model, frequently results in de-synchronization effect which most 3D watermarking methods are sensitive to and normally regard it as an attack. A novel pose-invariant blind watermarking algorithm for 3D triangular mesh model is proposed in this paper to resist the pose changing attack. In this algorithm, the geodesic distance of each vertex pair on the 3D mesh model to be watermarked is calculated, and then this model is segmented into meaningful pose-invariant components based on the calculation result. Each meaningful component is further divided into non-overlapping cells according to the distribution of geodesic distances from its vertices to the cutting boundary. Finally the defined watermark is embedded into these cells by adjusting the mean value of geodesic distances in order to achieve the invariance of pose changing, and later detection follows the same concept. Our experimental result demonstrates that the proposed algorithm has the ability to resist the attacks of pose changing, similarity transformations, vertex reordering, local deformation, cropping, simplification, additive random noise, and smoothing.

In this paper, a feature-based robust digital image watermarking algorithm is proposed to achieve the goal of image authentication and protection simultaneously. The Hessian-Affine feature detector is at first adopted to extract characteristic regions of an image in the proposed watermark embedding scheme. Then the copyright watermark is embedded into these characteristic regions according to the local orientation of each pixel. Moreover, the remainder regions are applied for image authentication by using block-wise fragile watermarking method. Similarly, the proposed watermark detection scheme follows the above procedure to extract the copyright watermark and the authentic information independently and blindly from watermarked images. Various attacks are also applied to the watermarked images in order to examine the robustness of our algorithm. The experimental results in this paper show that the proposed watermarking algorithm can resist most removal and geometric attacks. Besides, changes or modifications of an image will be reflected in our hidden watermarks.

In this paper a two-stage mode selection (TSMS) algorithm is presented to speed up the H.264/AVC video encoding process with rate distortion optimization (RDO). However, lots of additional computing power is required and this makes the realization of H.264/RDO in a resource-limited system very difficult. The proposed TSMS employs a two-stage decision process: the first stage is to predict some probable encoding modes according to the information when one encodes the preceding macroblocks and video frames. The second stage refines the decision with techniques based on Baye's probability rule and Back-Propagation neural network (BPN). According to the experiment results, over 50% of the computation time is reduced with very slight loss in peak signal-to-noise ratio (PSNR) and a slightly increment in bit rate when TSMS is applied. The TSMS is even faster than the encoding program with running RDO part. All programs are based on the H.264/AVC standard reference software (JM 9.2).

A novel neural network based method for halftoning and inverse halftoning of digital images is presented. We first start from inverse half-toning of images produced from error diffusion methods using an RBF network plus an MLP network. The restored contone images have had good quality already. Then, an SLP neural network is used to refine the halftoning processing and the training process of the inverse half-toning network is also involved. The combined training procedure produces half-tone images and the corresponding continuous tone images at the same time. It is found that these contone images have even better PSNR performance. Furthermore, the resulted half-tone images are visually sharper and clearer, too. The proposed inverse half-toning method is also compared to the well-known LUT method.