Focus measures are widely used to assess image clarity in various fields, such as photography and computer vision. However, many existing focus measures face challenges in balancing noise robustness and measurement capability. In this letter, a novel focus measure called Variance of Tight Framelet Feature (VTFF) is proposed to address this challenge. VTFF leverages the advantages of tight framelet features and variance information in feature maps to provide a robust and accurate assessment of image focus. Experimental results on both synthetic and real-world data demonstrate its superior performance compared to recent focus measures in measurement capability, noise robustness, and real-time performance.

Gait recognition has been greatly improved by deep learning and can achieve a relative high accuracy. The advances depend on the data size of gait. However, due to public concerns on privacy and regulations and laws from different countries, it is very difficult and almost impossible to collect a huge centralized gait database for algorithm training. Federated learning is a distributed machine learning technique for privacy-preserving, and can help to solve the problem. We propose a federated gait recognition benchmark, FedGait, to train algorithms using distributed gait data. It is the first benchmark on gait recognition to the best of our knowledge. FedGait can utilizes the gait videos available on multiple clients to learn a robust and generalized model. Based on the real-world gait scenarios, we introduce two federated gait recognition scenarios: institution-based scenario (IBS) and device-based scenario (DBS). Compared with centralized training, federated learning will encounter more serious heterogeneous data and data imbalance problems. We employ four popular databases for experiments, CASIA-B, CASIA-E, ReSGait and OU-MVLP, are involved in FedGait to investigate the problems in federated learning. We hope FedGait is a good start to solve data privacy problem in gait recognition.

Gait recognition is an effective way to identify a person due to its non-contact and long-distance acquisition. In addition, the length of human limbs and the motion pattern of human from human skeletons have been proved to be effective features for gait recognition. However, the length of human limbs and motion pattern are calculated through human prior knowledge, more important or detailed information may be missing. Our method proposes to obtain the dynamic information and static information from human skeletons through disentanglement learning. In the experiments, it has been shown that the features extracted by our method are effective.

Many studies have shown that gait recognition can be used to identify humans at a long distance, with promising results on current datasets. However, those datasets are collected under controlled situations and predefined conditions, which limits the extrapolation of the results to unconstrained situations in which the subjects walk freely in scenes. To cover this gap, we release a novel real-scene gait dataset (ReSGait), which is the first dataset collected in unconstrained scenarios with freely moving subjects and not controlled environmental parameters. Overall, our dataset is composed of 172 subjects and 870 video sequences, recorded over 15 months. Video sequences are labeled with gender, clothing, carrying conditions, taken walking route, and whether mobile phones were used or not. Therefore, the main characteristics of our dataset that differentiate it from other datasets are as follows: (i) uncontrolled real-life scenes and (ii) long recording time. Finally, we empirically assess the difficulty of the proposed dataset by evaluating state-of-the-art gait approaches for silhouette and pose modalities. The results reveal an accuracy of less than 35%, showing the inherent level of difficulty of our dataset compared to other current datasets, in which accuracies are higher than 90%. Thus, our proposed dataset establishes a new level of difficulty in the gait recognition problem, much closer to real life.

Gait is a popular biometric pattern used for identifying people based on their way of walking. Traditionally, gait recognition approaches based on deep learning are trained using the whole training dataset. In fact, if new data (classes, view-points, walking conditions, etc.) need to be included, it is necessary to re-train again the model with old and new data samples. In this paper, we propose iLGaCo, the first incremental learning approach of covariate factors for gait recognition, where the deep model can be updated with new information without re-training it from scratch by using the whole dataset. Instead, our approach performs a shorter training process with the new data and a small subset of previous samples. This way, our model learns new information while retaining previous knowledge. We evaluate iLGaCo on CASIA-B dataset in two incremental ways: adding new view-points and adding new walking conditions. In both cases, our results are close to the classical `training-from-scratch' approach, obtaining a marginal drop in accuracy ranging from 0.2% to 1.2%, what shows the efficacy of our approach. In addition, the comparison of iLGaCo with other incremental learning methods, such as LwF and iCarl, shows a significant improvement in accuracy, between 6% and 15% depending on the experiment.

Motivation: The identification of repetitive elements is important in genome assembly and phylogenetic analyses. The existing de novo repeat identification methods exploiting the use of short reads are impotent in identifying long repeats. Since long reads are more likely to cover repeat regions completely, using long reads is more favorable for recognizing long repeats. Results: In this study, we propose a novel de novo repeat elements identification method namely RepLong based on PacBio long reads. Given that the reads mapped to the repeat regions are highly overlapped with each other, the identification of repeat elements is equivalent to the discovery of consensus overlaps between reads, which can be further cast into a community detection problem in the network of read overlaps. In RepLong, we first construct a network of read overlaps based on pair-wise alignment of the reads, where each vertex indicates a read and an edge indicates a substantial overlap between the corresponding two reads. Secondly, the communities whose intra connectivity is greater than the inter connectivity are extracted based on network modularity optimization. Finally, representative reads in each community are extracted to form the repeat library. Comparison studies on Drosophila melanogaster and human long read sequencing data with genome-based and short-read-based methods demonstrate the efficiency of RepLong in identifying long repeats. RepLong can handle lower coverage data and serve as a complementary solution to the existing methods to promote the repeat identification performance on long-read sequencing data. Availability and implementation: The software of RepLong is freely available at https://github.com/ruiguo-bio/replong. Contact: [email protected] or [email protected]. Supplementary information: Supplementary data are available at Bioinformatics online.

In landmark-based image registration, estimation of transformation based on radial basis functions (RBFs) expansions has been successfully utilized in many applications. A novel landmark-based image registration method regularized by l1 and l2 norm is proposed in this paper to estimate transformations based on corresponding landmarks. The compact supported radial basis functions (CSRBFs) are utilized in our method. To estimate the CSRBFs coefficients of transformations, we construct a linear model and respectively regularize the elastic and affine deformation coefficients by l1 and l2 norm. Experiments show that the transformations estimated by our method are robust to noised correspondences of landmarks, the bending energy of transformations is less and topology of the deformation field can be preserved better than existing other methods.

We propose a new method to solve the point matching problem by l1 regularization. The non-rigid transformation function based on compact support radial basis functions (CSRBF) is represented by the linear system with respect to its coefficients. The transformation function is estimated by the proposed sparse optimization model with regularizing the CSRBF coefficients by l1 norm and the affine coefficients by the square of l2 norm. The optimization model for linear problem of transformation function can be efficiently solved by a fast iterative shrinkage-thresholding algorithm (FISTA) to accelerate the convergence speed of iterative procedure. Experiments on simulated point sets and lung datasets show that our method by l1 regularization obtains accurate registration results and is robust to estimate the correspondence and the transformation between two point sets in the presence of noise and outlier.

Pedestrian detection based on the framelet features in noisy depth images is investigated in this paper. For capturing the local features and attenuating the effects of noise in depth images, a features optimization model is proposed to adaptively select the framelet features for classification. The selected framelet features extracted by the model and SVM with a linear kernel is adopted as the feature and classifier, respectively. The proposed framelet features under a tight and redundant system can preserve the shape information while reducing the impact of noise. Experimental results also show that the proposed method based on framelet features can achieve a great improvement in noisy depth images, and the improvement is over one order of magnitude than HDD and HOG.

In this paper, we propose an image inpainting optimization model whose objective function is a smoothed l(1) norm of the weighted nondecimated discrete cosine transform (DCT) coefficients of the underlying image. By identifying the objective function of the proposed model as a sum of a differentiable term and a nondifferentiable term, we present a basic algorithm inspired by Beck and Teboulle's recent work on the model. Based on this basic algorithm, we propose an automatic way to determine the weights involved in the model and update them in each iteration. The DCT as an orthogonal transform is used in various applications. We view the rows of a DCT matrix as the filters associated with a multiresolution analysis. Nondecimated wavelet transforms with these filters are explored in order to analyze the images to be inpainted. Our numerical experiments verify that under the proposed framework, the filters from a DCT matrix demonstrate promise for the task of image inpainting.

This paper studies a problem of image restoration that observed images are contaminated by Gaussian and impulse noise. Existing methods for this problem in the literature are based on minimizing an objective functional having the l(1) fidelity term and the Mumford-Shah regularizer. We present an algorithm on this problem by minimizing a new objective functional. The proposed functional has a content-dependent fidelity term which assimilates the strength of fidelity terms measured by the l(1) and l(2) norms. The regularizer in the functional is formed by the l(1) norm of tight framelet coefficients of the underlying image. The selected tight framelet filters are able to extract geometric features of images. We then propose an iterative framelet-based approximation/sparsity deblurring algorithm (IFASDA) for the proposed functional. Parameters in IFASDA are adaptively varying at each iteration and are determined automatically. In this sense, IFASDA is a parameter-free algorithm. This advantage makes the algorithm more attractive and practical. The effectiveness of IFASDA is experimentally illustrated on problems of image deblurring with Gaussian and impulse noise. Improvements in both PSNR and visual quality of IFASDA over a typical existing method are demonstrated. In addition, Fast_IFASDA, an accelerated algorithm of IFASDA, is also developed.

We present a variational approach to obtain high-resolution images from multiframe low-resolution video stills. The objective functional for the variational approach consists of a data fidelity term and a regularizer. The fidelity term is formed by adaptively mimickingl1andl2norms. The regularization uses thel1norm of the framelet coefficients of a high-resolution image with a geometric tight framelet system constructed in this paper. The tight framelet system has abilities to detect multi-orientation and multi-order variations of an image. A two-phase iterative method for super-resolution reconstruction is proposed to construct a high-resolution image. The first phase is to get an approximation of the solution (i.e., the ideal image) using the steepest descent method. The second phase is to enhance the sparsity of the approximate solution by using the soft thresholding operator with variable thresholding parameters. Numerical results based on both synthetic data and real videos show that our algorithm is efficient in terms of removing visual artifacts and preserving edges in restored images.