Few-shot Class-incremental Pill Recognition (FSCIPR) aims to develop an automatic pill recognition system that requires only a few training data and can continuously adapt to new classes, providing technical support for applications in hospitals, portable apps, and assistance for visually impaired individuals. This task faces three core challenges: overfitting, fine-grained classification problems, and catastrophic forgetting. We propose the Well-Prepared Few-shot Class-incremental Learning (WP-FSCIL) framework, which addresses overfitting through a parameter-freezing strategy, enhances the robustness and discriminative power of backbone features with Center-Triplet (CT) loss and supervised contrastive loss for fine-grained classification, and alleviates catastrophic forgetting using a multi-dimensional Knowledge Distillation (KD) strategy based on flexible Pseudo-feature Synthesis (PFS). By flexibly synthesizing any number of old-class features, the PFS strategy resolves the issue of insufficient samples in the KD process, enabling Response-based KD (KD1) and Relation-based KD (KD2) to comprehensively preserve old knowledge. The effectiveness of WP-FSCIL has been validated through experiments conducted on two publicly available pill datasets. These experiments show that WP-FSCIL outperforms existing state-of-the-art methods, demonstrating its superior performance.

Environmental Microorganisms (EMs) serve as vital bioindicators for monitoring environmental pollution monitoring. However, their classification is challenged by severe class imbalance and a lack of labeled data. Precise identification of EMs is essential for timely ecological risk assessment, yet conventional methods struggle in few-shot learning (FSL) contexts due to poor generalization and overfitting. To overcome these challenges, we introduce a novel Cross-stage Multi-branch Deep Feature Extraction (CMDFE) framework, which integrates deep feature extraction (Re) with a lightweight multi-branch module (C2f). This design improves feature diversity and mitigates overfitting, while maintaining a favorable balance between accuracy and computational efficiency. Experimental results on the EMDs7 dataset demonstrate the effectiveness of CMDFE, achieving$\mathbf{6 8. 4 0 \%}$accuracy in 1-shot and$\mathbf{7 7. 0 0 \%}$in 5-shot tasks, surpassing most baseline methods. Additionally, CMDFE significantly reduces memory usage, making it well-suited for deployment in resource-constrained environments. Overall, CMDFE offers a robust and scalable approach for few-shot EM classification and holds promise for intelligent environmental monitoring applications.

This study applies a Few-shot learning algorithm (Met-FSNet) to the multi-target semantic segmentation of renal parenchyma, perirenal fat, liver, spleen, gallbladder, subcutaneous fat, ribs, and vertebrae in CT images. The proposed Fewshot segmentation model incorporates both a ST module and a Fusion module, which enhance the model's spatial adaptability and ability to integrate multi-scale features, resulting in improved robustness. The model demonstrates strong generalization capabilities, performing well not only for parenchymal organs (kidney, liver, and spleen) but also for hollow organs (gallbladder), bony structures (ribs and vertebrae), and adipose tissues (perirenal and subcutaneous fat). Experiments were conducted using a private multi-target segmentation dataset, achieving excellent segmentation results across all eight categories. Subcutaneous fat yielded the highest IOU of 93 %, followed by the liver at 91.1 %. All other categories showed improved IOU metrics compared to the baseline network. This work holds significant clinical value for physicians in the early diagnosis of metabolic syndrome.

This study proposes an explainable few-shot learning framework for cervical cancer nest classification, addressing challenges of limited samples and dense cellular overlap. Integrating Model-Agnostic Meta-Learning (MAML) with a dual-attention feature extraction module (spatial-focused CBAM and frequency-domain FECAM), our method achieves efficient pathological feature representation in Euclidean space. Evaluated on 3,931 cervical cytopathology images (2,000 normal; 1,931 abnormal across four subtypes), the framework attained 92.3 %,$\mathbf{8 7. 5 0 \%, ~} \mathbf{9 7. 8 \%}$, and$\mathbf{9 1. 2 \%}$on average in the normal category for Precision, Recall, Specificity, F1-Score, and Accuracy respectively. The average accuracy and variance of the four replicate experiments is$\mathbf{0. 6 9 7}$and$6.02 \mathrm{e}-5$, respectively. It significantly outperformed 18 benchmark models. Meanwhile, the mathematical formalization validates model interpretability, while the dualattention mechanism enhances discriminative feature capture of nuclear abnormalities.

Metabolic syndrome (MetS) is a group of metabolic disorders with insulin resistance at its core, with a high global prevalence, and early screening is crucial for preventing cardiovascular disease. Currently, clinical diagnosis relies on invasive biochemical testing, which is expensive and patient compliance is poor. Infrared thermography (IRT) has the potential to screen for MetS as a non-invasive functional imaging technique, but high-resolution devices are expensive and low-cost devices have low resolution. In this paper, a method combining multimodal imaging and super-resolution reconstruction is proposed to construct a synchronous acquisition system containing low-resolution infrared, high-resolution infrared and visible light images. At the algorithm level, the system evaluates SwinIR and lightweight SRVGG networks to reconstruct a low-resolution image of$32 \times 24$pixels to$256 \times 192$pixels. The experimental results show that SwinIR has the best performance in PSNR (30.51 dB) and SSIM (0.872), and SRVGG is still significantly better than the traditional interpolation method despite its light weight (79.1 KB). This study provides an effective technical path to overcome the cost and resolution limitations of infrared equipment, which is helpful to promote the early screening and prevention and control of MetS at the community and family levels.

Histopathological image classification in colorectal cancer (CRC) is hindered by class imbalance, where limited data for some classes and overlapping features lead to degraded performance. Traditional methods struggle to adequately augment minority classes or resolve these uncertainties. To alleviate these problems, we propose a novel data synthesis and adaptive loss-based histopathological image classification framework that integrates two components: 1) a global context generative adversarial network (GCGAN) to generate realistic images for under-represented classes, thereby enriching dataset diversity; 2) an adaptive focal loss (AFL) to dynamically adjust the loss function and mitigate feature overlap. Extensive experiments on the CRC dataset demonstrate the effectiveness of our approach, outperforming other state-of-the-art (SOTA) methods. Additionally, combining GCGAN and AFL benefits these SOTA methods, with performance improvements on cervical (CC) and breast cancer (BC) datasets.

As the demand for Microscopic Image Analysis (MIA) increases, object detection in deep learning has gradually become a popular research area. Considering the high cost of microscopic image acquisition and dataset preparation, training data is often insufficient and imbalanced. Few-shot Object Detection (FSOD) offers a new approach. However, current FSOD methods have not fully utilized color and texture characteristics when processing microscopic images, which still have potential performance that has not been explored. Therefore, we propose Frequency-domain Feature Enhanced Faster-RCNN (FFEFRCN), which includes a frequency-enhanced feature pyramid network, gradient decoupling layers, and a channel attention calibration block. We conducted benchmark experiments on the Environmental Microorganism Image Dataset Seventh Version (EMDS-7) and extended experiments on the Sperm Videos and Images Analysis (SVIA). Experimental results show that our model achieved superior performance. On EMDS-7, we achieved an mAP50 and nAP50 of 56.06 % and 41.00 % at 1-shot, and 73.39 % and 69.20 % respectively at 8 -shots. On SVIA, we achieved an AP50 of 96.49 % for sperm and 70.21 % for impurity.

This paper presents a fully automatic method, BladeView, for drone-based wind turbine blade inspection using an Unmanned Aerial Vehicle (UAV). With the need for highly efficient blade inspection coupled with the rapid increase of wind turbines, existing methods provide limited automation on wind turbine parameter estimation, full blade coverage, and safety control. We introduce an Automatic Parameter Calculation (APC) algorithm and an Automatic Flight System (AFS) in BladeView to compute wind turbine parameters and inspection paths, respectively. Leveraging triangulation and linear fitting integration techniques, the APC automatically calculates the wind turbine parameters and estimates the relative angle and position between a drone and the turbine. Furthermore, with dynamic path finding and B-spline optimization, the AFS plans a path covering 3 blades within specified flight corridors, in compliance with the turbine parameters obtained from APC. Thus, the proposed BladeView can properly ensure an inspection’s automation, coverage, safety, and smoothness. The efficiency and usability of BladeView are validated through 100,000 flight simulations in the Gazebo simulation environment and 9,239 field runs at various wind farms, including offshore, near-shore, deserts, mountainous areas, farmlands, and suburbs.Note to Practitioners—The proposed BladeView is distinguished in three aspects: (1) It automatically adapts wind turbines with varying geometric properties and physical locations relative to the take-off point. (2) It dramatically improves the quality of collected data with optimal UAV speed and flight corridors. (3) It thoroughly covers all three blades of a Horizontal-Axis Wind Turbine (HAWT), including regions where defects frequently occur. Thus, BladeView is more efficient and robust than existing UAV-based methods for blade inspection, with only around 25 minutes per HAWT. Moreover, it does not require experienced pilots to fly the UAV and manual interventions are rarely needed. Extensive simulation and real-world experiments demonstrate the efficiency and usability of BladeView in various on- and offshore wind farms.

Denoising Diffusion Probabilistic Models (DDPMs) have attracted much attention as a powerful generative model and have been widely used in the field of computer vision. In the field of medical images, the noise contained in low-dose images often poses a challenge to deep learning methods, so we propose a Conjugate and Hybrid Diffusion Model (CH-DDPMs) for the task of low-dose fat CT image segmentation. CH-DDPMs is a kind of DDPMs integrated with a DnCNN module. It solves the problem of the noise interference in the low-dose fat CT images. CH-DDPMs achieves better results than other models in images with different noise levels, which can help doctors diagnose MetS more effectively.

The accurate detection of sperms and impurities is a very challenging task, facing problems such as the small size of targets, indefinite target morphologies, low contrast and resolution of the video, and similarity of sperms and impurities. So far, the detection of sperms and impurities still largely relies on the traditional image processing and detection techniques which only yield limited performance and often require manual intervention in the detection process, thus unfavorably escalating the time cost and injecting the subjective bias into the analysis. Encouraged by the success of deep learning methods in numerous object detection tasks, here we report a deep learning network: ACTIVE based on Double Branch Feature Extraction Network (DBFEN) and Cross-conjugate Feature Pyramid Network (CCFPN). DBFEN extracts visual features from tiny objects with a double branch structure, and CCFPN fuses the features extracted by DBFEN to enhance the description of the position and high-level semantic information. Our work is the pioneer of introducing deep learning approaches to the detection of sperms and impurities. Experiments show that the highest AP50of the sperm and impurity detection is 91.13% and 59.64%, which lead its competitors by a substantial margin and establish the state-of-the-art results in this problem. Code is available for readers’ free evaluation at https://github.com/anheqiao-neu/ACTIVE.

Diabetes is one of the risk factors for metabolic syndrome and leads to an increased risk of cardiovascular events. Epicardial fat is located between the myocardium and the pericardium, directly contacting the coronary arteries, and is therefore closely related to cardiovascular events. We developed a method for segmenting epicardial fat, which achieved a DICE score of 0.6811 on the validation set. Based on this, we explored the correlation between epicardial fat, cardiac structure, and cardiac ejection function in diabetic patients. The results show that epicardial fat first alters the cardiac structure in diabetic patients, subsequently affecting their cardiac ejection function.

Recently, the development of artificial intelligence technologies has accelerated advancements in threshold segmentation techniques based on intelligent optimization algorithms, which have demonstrated good performance across various types of medical images. However, research specifically focused on the segmentation of Prostate Cancer Pathology Image (PCPI) remains limited. To fill this gap and improve the efficiency of prostate cancer diagnosis, this study proposes a threshold segmentation scheme based on intelligent optimization algorithms, MOLMVO-Rényi, specifically designed for PCPI. This scheme first enhances the optimization capabilities of the multi-verse optimizer algorithm using multiple strategies (multi-population topological structure and orthogonal learning) to propose MOLMVO. Then it utilizes MOLMVO to optimize the thresholds for image segmentation, aiming to maximize the Rényi entropy of the segmented images. Experimental validation on 67 privately collected PCPIs indicates that the proposed scheme shows significant competitive advantages, providing strong support for the early diagnosis and treatment of prostate cancer.

This study proposes a Generative Adversarial Network (GAN)-based data augmentation method to tackle the class imbalance issue in histopathological image classification. The proposed GAN generates high-quality images for minority classes, improving data diversity and classification performance. By integrating Global Context Attention (GCA) and UpBlock-CompRes modules, the GAN generates images that closely resemble real data, achieving superior results in FID, IS, PSNR, and SSIM compared to FastGAN. Experimental results demonstrate that the proposed GAN outperforms both the baseline without data augmentation and FastGAN, with noticeable improvements across key metrics. These enhancements highlight the effectiveness of our GAN in generating high-quality images and achieving more balanced classification results.

This work proposes a Few-shot network employed for Perirenal Fat and Renal Parenchyma Semantic Segmentation (PRS-Net) in Diabetic Kidney Disease (DKD). PRS-Net integrates ST module and Fusion module. These two modules allow the model to process CT images with different spatial distributions and fuse multi-scale features, thereby enhancing performance in image segmentation. Utilizing the Perirenal Fat and Renal Parenchyma Dataset for semantic segmentation, PRS-Net achieves a mean intersection over union of 60.22% on test set, achieving superior performance compared to the other models. PRS-Net has clinical significance for early DKD diagnosis.

Metabolic Syndrome (MetS) is a prevalent condition associated with an increased risk of cardiovascular diseases, characterized by high blood pressure, hyperglycemia, and dyslipidemia. These risk factors not only exacerbate cardiovascular conditions but also impair immune function. Timely detection and prevention of MetS are imperative to mitigate these health risks. Recent advancements indicate various diagnostic approaches, including electrochemical biomarker detection, muscle mass to visceral fat ratio assessments, and anthropometric indices such as the body roundness index. Moreover, multimodality imaging techniques have become essential tools in comprehensive evaluation of MetS. This study introduces a cost-effective infrared thermal imager designed for MetS ordinary monitoring. Through simulation experiments involving 20 participants, 400 images of samples are collected and analyzed. The results demonstrate significant differences in thermal images between negative and positive samples. This innovative method could potentially offer a cost-effective and non-invasive tool for MetS monitoring.

Cervical cancer poses a significant threat to the health of women worldwide. Cervical cytopathology screening is an effective method for diagnosing cervical cancer. However, manual screening is time-consuming and prone to errors. The advent of automatic Computer-Aided Diagnosis (CAD) systems based on deep learning addresses this problem, however training these models requires large amounts of labeled data, which may not always be available. This paper proposes a Few-Shot Learning (FSL) framework called FSL-DSC to perform cervical cell classification tasks on small dataset. FSL-DSC first proposes inner loop learning and outer loop learning for individual tasks and overall parameter updates respectively, then a depthwise separable module is designed to further enhance the performance of the model. Among three repeated experiments, the FSL-DSC framework achieves an average accuracy of 83.34%, which shows the effectiveness and potential of the proposed FSL-DSC in the field of cervical image classification and few-shot tasks.

Smartphone-based indoor positioning and navigation remains a challenging task, as specialized technologies such as ultra-wideband (UWB) or Wi-Fi fine-time measurement are still niche and supported by only a few flagship smartphones. Therefore, standard technologies based on RSSI measurements, mainly Bluetooth Low Energy (BLE) and Wi-Fi, are used to obtain absolute positioning information of pedestrians inside buildings. Sensor fusion methods combine this with relative information from modeling human movement using sensor data provided by the smartphone’s IMU. It is also common practice to restrict this movement to the actual accessible areas of the building (e.g. restricting moving through walls), using spatial models based on the building’s floor plan. Without further assumptions, this complexity inevitably leads to a non-linear and non-Gaussian state space model. A common tool for (position) estimation in such scenarios is the broad class of particle filters. However, the use of such spatial constraints accelerates the well-known problem of sample impoverishment, which in the worst case can lead to the particle filter completely losing track, getting stuck and never recovering. This work begins with a brief presentation of an award-winning Indoor Positioning System (IPS) derived from previous work. Based on this, we present several approaches using Particle Distribution Optimization (PDO) that attempt to solve the impoverishment problem and ultimately lead to better overall positioning results. In the experiments, we compare them in two different buildings under realistic conditions and discuss the results in detail.

Acquiring accurate ground truth or fingerprint data for indoor localization typically demands specialized hardware and considerable effort. This paper introduces three post-processing methods to simplify the recording of such data within certain limitations. The methods can be categorized into fixed-interval smoothing where the complete recording data is available for the optimal state estimation and there is no live processing requirement. Our methods are based on a particle filter system, employing the CONDENSATION algorithm to estimate the hidden state. Particles represent positions and headings in 3D space. They are moved and evaluated using pedestrian dead reckoning based on IMU measurements and a map of the building only. The development and application of these methods is presented in this paper, demonstrating their effectiveness in reducing the effort required for accurate ground truth and fingerprint data recording.

Abstract Skeleton Ground Truth (GT) is critical to the success of supervised skeleton extraction methods, especially with the popularity of deep learning techniques. Furthermore, we see skeleton GTs used not only for training skeleton detectors with Convolutional Neural Networks (CNN), but also for evaluating skeleton-related pruning and matching algorithms. However, most existing shape and image datasets suffer from the lack of skeleton GT and inconsistency of GT standards. As a result, it is difficult to evaluate and reproduce CNN-based skeleton detectors and algorithms on a fair basis. In this paper, we present a heuristic strategy for object skeleton GT extraction in binary shapes and natural images. Our strategy is built on an extended theory of diagnosticity hypothesis, which enables encoding human-in-the-loop GT extraction based on clues from the target’s context, simplicity, and completeness. Using this strategy, we developed a tool, SkeView, to generate skeleton GT of 17 existing shape and image datasets. The GTs are then structurally evaluated with representative methods to build viable baselines for fair comparisons. Experiments demonstrate that GTs generated by our strategy yield promising quality with respect to standard consistency, and also provide a balance between simplicity and completeness.

In the medical field, on the one hand, data mining can effectively establish evaluation models to supplement gold standards; on the other hand, it can guide the direction of scientific research by establishing connections between knowledge. Radiology images and pathological images are considered to be the most suitable medical data for data mining due to their large amount of information. Endometrial carcinoma is a common malignant tumor in women, and TP53 mutation status is an important factor affecting the occurrence and development of tumors. In this study, we propose a neural network structure based on multi-modal medical data that can predict TP53 mutations in endometrial carcinoma, with an accuracy of 86.21% in test set. Then, we clustered TP53-related deep learning features, and we believe that there is heterogeneity in TP53-related deep learning features.

For esophageal cancer immunotherapy, Programmed Death Ligand-l (PD-LI) is considered a predictive biomarker. However, immunehistochemistry (IHC) methods used to quantify PD-LI are challenged by high cost, time and variability. In contrast, hematoxylin and eosin (H&E) staining is a reliable method commonly used in cancer diagnosis. By employing advanced deep learning techniques, this study demonstrates the feasibility of predicting PD-LI expression from H&E stained images. With the help of pathologists, a dataset is constructed to evaluate the validity of PD-LI prediction in esophageal cancer by H&E using the FusedNet model. In 227 patients, PD-LI status is systematically predicted. Consistent prediction performance is demonstrated through validation of the validation set, proving that the system can be used as a decision support and quality assurance system in clinical practice.

Background and purpose: Skin tumours have become one of the most common diseases worldwide. While benign ones are not usually a threat to human health, malignant ones can develop into skin cancer and become life-threatening if left untreated. Early detection of the disease is important for the treatment of patients with skin tumours and dermoscopy is the most effective means of diagnosing skin tumours. However, the complexity of skin tumour cells makes the diagnosis somewhat erroneous for doctors. Therefore, a dermoscopic classification network based on deep learning and computer-aided diagnostic techniques is needed to obtain a high diagnostic accuracy rate for skin tumours. Methods: In this paper, Deep-skin, a model for dermoscopic image classification is proposed, which is based on both attention mechanism and ensemble learning. Considering the characteristics of dermoscopic images, embedding different attention mechanisms on top of Inception-V3 has been suggested to obtain more potential features. We then improve the classification performance by late fusion of the different models. To demonstrate the effectiveness of Deep-skin, experiments and evaluations are performed on the publicly available dataset Skin Cancer: Malignant vs. Benign and compare the performance of Deep-skin with other classification models. Results: The experimental results indicate that Deep-skin performs well on the dataset in comparison to other models, achieving a maximum accuracy of 87.8%.Conclusion: In this paper, the Deep-skin model is proposed for the classification of dermoscopic images and has shown better performance. In the future, we intend to investigate better classification models for automatic diagnosis of skin tumours. Such models can potentially assist physicians and patients in clinical settings.

The detection of environmental microorganisms is always a difficult task, e specially when the multi-scale environment is complex. For tiny objects in microscopic images, current detection methods face the challenge of accurate identification and localization. In contrast, we propose a convolutional neural network (ECA-RetinaNet) for microscopic object detection of which underlying dataset is a high-quality EMDS-7 dataset. The accuracy of ECA-RetinaNet is high, with a high mean Average Precision (mAP) value of 81.42% in the Environmental Microorganisms (EMS) detection task. Its accuracy has been higher than that of the two-stage object detection network.

The involvement of a multitude of parameters adds to the complexity of modeling a flood. However, floods are among the most destructive of natural disasters and therefore, flood forecasting is one of the key priorities of hydrology. Flood forecasting goes a long way to minimize the loss of lives as well as economic losses. Furthermore, proper modeling of floods can contribute immensely towards future risk reduction and the introduction of necessary policies. At present, the application of machine learning in river and flood analysis has dramatically increased among hydrologists. In this research, we propose a location (District) independent flood prediction model (Random Forest-RF) of commercially significant rivers in Bangladesh. The data Imbalance problem is solved by synthetic minority oversampling of numerical and categorical (SMOTENC) data augmentation techniques. Our results show that the proposed framework outperforms the previously reported results by up to 9%. To the best of our knowledge, our proposed flood prediction framework achieved the best performance in terms of all evaluation matrices on the specific dataset.

This work presents SIMUL, a new dataset consisting of 550 minutes of annotated motion data from six synchronized IMUs placed consistently at the same strategically chosen positions on the bodies of 32 participants. With a focus on indoor localization, the selection fell on hand, feet, and trouser pockets. Due to the sensor’s synchronization, this selection allows, for example, to label the data recorded in the hand based on the data captured at the feet. For a better generalizability, the dataset was recorded freestyle under many different environmental conditions and walking speeds. Thus, indoors, numerous floor coverings such as stone, wood, carpet and PVC, as well as stairs are included in the data. The outdoor recordings additionally contain uneven surfaces such as paving stones and slopes. The annotation of the participant’s currently performed activity additionally allows this dataset to be used for activity recognition.

Detecting moiré patterns in digital photographs is meaningful as it provides priors towards image quality evaluation and demoiréing tasks. In this paper, we present a simple yet efficient framework to extract moiré edge maps from images with moiré patterns. The framework includes a strategy for training triplet (natural image, moiré layer, and their synthetic mixture) generation, and a Moiré Pattern Detection Neural Network (MoireDet) for moiré edge map estimation. This strategy ensures consistent pixel-level alignments during training, accommodating characteristics of a diverse set of camera-captured screen images and real-world moiré patterns from natural images. The design of three encoders in MoireDet exploits both high-level contextual and low-level structural features of various moiré patterns. Through comprehensive experiments, we demonstrate the advantages of MoireDet: better identification precision of moiré images on two datasets, and a marked improvement over state-of-the-art demoiréing methods.

Radio frequency ranging protocols enable a device to estimate the distance to another device, based on signal propagation time. In theory, ranging protocols are promising for indoor localization as one can obtain the position of the pedestrian directly from the ranging results if the access point positions are known. However, in practice, indoor scenarios still pose a challenging problem as the observed distances vary greatly due to non-line-of-sight signal paths, delayed signal propagation, and general hardware inaccuracies. The IEEE 802.11-2016 (formerly IEEE 802.11mc) standard defines a RF ranging protocol for Wi-Fi, namely Fine Timing Measurement (FTM). In order to improve the position estimate, a novel sensor model for FTM is derived from observed data. It is shown that the FTM error varies with the actual distance to the access point. Within this work, different parameter sets are estimated from the observed data for skew normal distributions, depending on the actual distance. For these parameters, low-order polynomials are then fitted to obtain the distribution parameters as functions of the actual distance. Furthermore, a particle filter is described and evaluated in an industrial scenario using cheap Espressif ESP32-S2 IoT FTM access points in the 2.4 GHz band. The filter combines map information, Pedestrian Dead Reckoning, and our novel FTM sensor model to estimate the pedestrian's position in the building. Finally, the localization result of the particle filter is compared to another promising radio frequency ranging method: ultra-wideband.

Particle filters are a commonly used technique for sensor fusion in indoor localization use-cases. Multiple strategies exist, that control when the particle filter is updated and with which data. We take a look at three commonly used update strategies. The first runs a full particle filter update for every incoming measurement, the second uses a fixed-interval update rate and the third is triggered by events, such as detected steps. All of these strategies have different advantages and disadvantages. The first strategy, for example, has the problem that steps are recognized only after they have been completed - which makes for a constant temporal discrepancy between the proposal distribution and the measurements evaluated on top. Due to the configured delay, the fixed-interval strategy, in comparison, can pre-date incoming step events to mitigate this discrepancy. In this paper we present PIPF as a novel approach to combine advantages of the fixed-interval update strategy with advantages of the strategy that runs a full update per measurement. This works by using the transition to calculate a trajectory, which is then followed during the evaluation. The proposal distribution in the form of particles is interpolated on this trajectory for every incoming measurement, which removes the temporal discrepancy between both. To evaluate PIPF's performance and characteristics at different configurations, we compare it to a conventional fixed-interval particle filter on a real-world indoor positioning scenario.

In this paper, we present a simple yet efficient framework, BlumNet, for extracting object skeletons in natural images and binary shapes. With the need for highly reliable skeletons in various multimedia applications, the proposed BlumNet is distinguished in three aspects: (1) The inception of graph decomposition and reconstruction strategies further simplifies the skeleton extraction task into a graph component detection problem, which significantly improves the accuracy and robustness of extracted skeletons. (2) The intuitive representation of each skeleton branch with multiple structured and overlapping line segments can effectively prevent the skeleton branch vanishing problem. (3) In comparison to traditional skeleton heatmaps, our approach directly outputs skeleton graphs, which is more feasible for real-world applications. Through comprehensive experiments, we demonstrate the advantages of BlumNet: significantly higher accuracy than the state-of-the-art AdaLSN (0.826 vs. 0.786) on the SK1491 dataset, a marked improvement in robustness on mixed object deformations, and also a state-of-the-art performance on binary shape datasets (e.g. 0.893 on the MPEG7 dataset).

This article introduces a novel approach to generate visually promising skeletons automatically without any manual tuning. In practice, it is challenging to extract promising skeletons directly using existing approaches. This is because they either cannot fully preserve shape features, or require manual intervention, such as boundary smoothing and skeleton pruning, to justify the eye-level view assumption. We propose an approach here that generates backbone and dense skeletons by shape input, and then extends the backbone branches via skeleton grafting from the dense skeleton to ensure a well-integrated output. Based on our evaluation, the generated skeletons best depict the shapes at levels that are similar to human perception. To evaluate and fully express the properties of the extracted skeletons, we introduce two potential functions within the high-order matching protocol to improve the accuracy of skeleton-based matching. These two functions fuse the similarities between skeleton graphs and geometrical relations characterized by multiple skeleton endpoints. Experiments on three high-order matching protocols show that the proposed potential functions can effectively reduce the number of incorrect matches.

A vision-based method for detecting the cracks in the concrete sleepers of the railway tracks will be introduced in this paper. The method is able to detect and partially classify the cracks of the concrete sleepers in two successive steps based on the image processing and pattern recognition techniques. The method has been implemented on the acquired image data frames followed by the analysis, experimental, comparison results and evaluation. The presented results are reasonable which indicates the goodness of the introduced method. The preliminary results of this work have been presented in [A. Delforouzi, A. H. Tabatabaei, M. H. Khan and M. Grzegorzek, A vision-based method for automatic crack detection in railway sleepers, in Kurzynski, M., Wozniak, M., Burduk, R. (eds.), Proceedings of the 10th International Conference on Computer Recognition Systems CORES 2017, Polanica Zdroj, Poland. CORES 2017. Advances in Intelligent Systems and Computing, Vol. 578 (Springer, Cham, 2018), pp. 130–139, doi: 10.1007/978-3-319-59162-9_14].

Emotion recognition is a increasingly popular topic because of its potential applications in the field of affective learning. It allows the development of systems able to adapt themselves to the users' emotional state to improve the learner's experience and learning. In this paper, we introduce a new biomedical multi-sensor platform for realtime acquisition of physiological data comprising Temperature, Electroencephalography (EEG), Electroocculography (EOG), Galvanic Skin Response (GSR), Heart Rate and Blood Oxygen Saturation. We describe experimental scenarios for the induction of emotions relevant in a context of affective learning (happiness, frustration, boredom) to build a set of emotionrelated data. We carry out a basic classification study by computing hand-crafted features on the time and frequency domains of signals, and training a Support-Vector-Machine (SVM) classifier to demonstrate the feasibility of our approach.

It is common practice to use a sample-based representation to solve problems having a probabilistic interpretation. In many real world scenarios one is then interested in finding a best estimate of the underlying problem, e.g. the position of a robot. This is often done by means of simple parametric point estimators, providing the sample statistics. However, in complex scenarios this frequently results in a poor representation, due to multimodal densities and limited sample sizes. Recovering the probability density function using a kernel density estimation yields a promising approach to solve the state estimation problem i. e. finding the “real” most probable state, but comes with high computational costs. Especially in time critical and time sequential scenarios, this turns out to be impractical. Therefore, this work uses techniques from digital signal processing in the context of estimation theory, to allow rapid computations of kernel density estimates. The gains in computational efficiency are realized by substituting the Gaussian filter with an approximate filter based on the box filter. Our approach outperforms other state of the art solutions, due to a fully linear complexity and a negligible overhead, even for small sample sets. Finally, our findings are evaluated and tested within a real world sensor fusion system.

This paper presents a novel cross-view gait recognition technique based on the spatiotemporal characteristics of human motion. We propose a deep fully-connected neural network with unsupervised learning which transfers the gait descriptors from multiple views to the single canonical view. The proposed non-linear network learns a single model for all videos captured from different viewpoints and finds a shared high-level virtual path to map them on a single canonical view. Therefore, the model does not require any labels or viewpoint information in the learning phase. The network is learned only once using the spatiotemporal motion features of the gait sequences from several viewpoints, later it is used to construct the cross-view gait descriptors for the gallery and the probe sets. The descriptors are classified using simple linear support vector machine. Experiments carried out on the benchmark cross-view gait dataset, CASIA-B, and comparisons with the state-of-the-art demonstrate that the proposed method outperforms the existing cross-view gait recognition algorithms.

The main goal of this paper is to provide an effective approach to quantify patterns of team managers so that they can be learned and compared by machines. Traditionally, team managers are analyzed and compared based on the data collected from surveys and questionnaires. Then, managers are usually represented by managerial roles/skills and analyzed semi-manually by human perception. However, it causes two methodological problems: 1) managerial roles and skills are usually isolated and only indirectly related to managerial actions and 2) the perception-based methods cannot provide detailed analysis results since human perception is abstract and lacks stability. In order to solve these problems, we propose a simple but powerful manager representation method which is general enough to cover most manager types. With this, team managers can be learned and compared my machines. Particularly, we represent a manager by managerial actions with flexible feature groups to improve the expandability and description power of the proposed representation model. For manager analysis, we introduce the first matching algorithm which not only returns robust and stable manager similarities but also details the matched parts among managerial action sequences. The efficiency of the proposed methods is substantiated by comparison experiments between machines and human perception.

For semantic analysis of activities and events in videos, it is important to capture the spatio-temporal relation among objects in the 3D space. In this paper, we present a probabilistic method that extracts 3D trajectories of objects from 2D videos, captured from a monocular moving camera. Compared with existing methods that rely on restrictive assumptions, we propose a method that can extract 3D trajectories with much less restriction by adopting new example-based techniques, which compensate the lack of information. Here, we estimate the focal length of the camera based on similar candidates, and use it to compute depths of detected objects. Contrary to other 3D trajectory extraction methods, our method is able to process videos taken from a stable camera as well as a non-calibrated moving camera without restrictions. For this, we modify Reversible Jump Markov Chain Monte Carlo particle filtering to be more suitable for camera odometry without relying on geometrical feature points. Moreover, our method decreases time consumption by reducing the number of object detections with keypoint matching. Finally, we evaluate our method on known data sets by showing the robustness of our system and demonstrating its efficiency in dealing with different kind of videos.

Biometric gait recognition has received substantial attention of researchers in the recent years due to its applications in numerous fields of computer vision, particularly in visual surveillance and monitoring systems. Most existing gait recognition algorithms solve the problem of person identification either by constructing a human body model based on various skeletal data characteristics such as joints positioning and their orientation, or use gait features, e.g., stride length, gait patterns and other shape templates. Such approaches require the extraction of the human-body's silhouette, contour, or skeleton from the images, and therefore their performance highly depends on the silhouette segmentation accuracy. In this paper, we propose a novel gait recognition algorithm which exploits spatiotemporal motion characteristics of a person, which does not need silhouette or skeleton extraction at all. The proposed algorithm computes a set of spatiotemporal features from the video sequences and uses them to generate a codebook. Fisher vector is used to encode the motion descriptors which are classified using linear Support Vector Machine (SVM). The proposed algorithm is evaluated on three benchmark gait datasets: TUM GAID, CASIA-B, and CASIA-C. It achieved excellent results on all datasets which demonstrate the effectiveness of the proposed algorithm.

In this paper we present an approach for real-time gesture recognition using exclusively 1D sensor data, based on the use of Particle Filters and Dynamic Time Warping Barycenter Averaging (DBA). In a training phase, sensor records of users performing different gestures are acquired. For each gesture, the associated sensor records are then processed by the DBA method to produce one average record called template gesture. Once trained, our system classifies one gesture performed in real-time, by computing -using particle filters-an estimation of its probability of belonging to each class, based on the comparison of the sensor values acquired in real-time to those of the template gestures. Our method is tested on the accelerometer data of the Multimodal Human Activities Dataset (MHAD) using the Leave-One-Out cross validation, and compared with state-of-the-art approaches (SVM, Neural Networks) adapted for real-time gesture recognition. It manages to achieve a 85.30% average accuracy and outperform the others, without the need to define hyper-parameters whose choice could be restrained by real-time implementation considerations.

In recent research, indoor localisation systems are often based upon a recursive state estimation using particle filtering. Within this context, sample impoverishment is a crucial problem causing the position estimation to lose track or get stuck within a demarcated area. The sample impoverishment problem can therefore be described as a too small particle diversity, unable to sample enough particles into proper regions of the dynamic system. Restrictive transition models, as they are used in indoor localisation, also enhance this effect significantly. However, an accurate position estimation requires a certain degree of focus and thus behaves contrary to the need of diversity. Therefore we propose a new method that is able to deal with the trade-off between the need of diversity and focus by deploying an interacting multiple model particle filter (IMMPF) for jump Markov non-linear systems. We combine two similar particle filters using a non-trivial Markov switching process, depending upon the Kullback-Leibler divergence and a Wi-Fi quality factor. The main benefit of this approach is an easy adaptation to other localisation approaches based on particle filters.

The task of object tracking in rectangular videos has been addressed in recent years by many researchers, where each method tries to propose a solution for a special challenge. Handling a variety of challenging situation of object tracking in 360-degree videos is still an unsolved problem and needs to be more considered. In the real world, the challenging situations include moving camera, high-resolution videos, background clutter, fast processing, in and out-of-plane rotation, small, articular and rigid objects, illumination variation and partial and full occlusion. In this paper, a new structure of SURF-based object tracking is proposed which uses a train-based matching to address the challenging object tracking in 360-degree videos. The proposed tracker is able to estimate out-of-plane rotation and occlusion during the tracking and adapt itself to handle it. Our experiments demonstrate the robustness of our tracker.

Vojta-therapy is a useful technique to treat the disorders in the central nervous and musculoskeletal system. During the therapy, a specific stimulation is given to the patients in order to cause the patient's body to perform certain reflexive pattern movements. The repetition of this stimulation ultimately brings forth the previously blocked connections between the spinal cord and brain, and after a few sessions, patients can perform these movements without any external stimulation. In this paper we proposed an automatic vision-based monitoring system for accurate therapy. We proposed an infant's (i.e., patient) detection and recognition of specific movements in his/her various body parts during the therapy process, using RGB-D data. First, A robust template matching based algorithm is exploited for infant's detection using his/her head location. Second, various features are computed to capture the movements of different body parts during the therapy. In the classification stage, a multi-class support vector machine (mSVM) is used to classify the accurate movements of infant during the therapy process, which ultimately reveals the correctness of the given treatment. The proposed algorithm is evaluated on our challenging dataset, which was collected in a children hospital. The detection and classification results show that the proposed method is highly useful to recognize the correct movement pattern either in hospital or in-home therapy systems.

A vast amount of toxicological data can be obtained from feature analysis of cells treated in vitro. However, this requires microscopic image segmentation of cells. To this end, we propose a new strategy, namely Supervised Normalized Cut Segmentation (SNCS), to segment cells that partially overlap and have a large amount of curved edges. SNCS approach is a machine learning based method, where loosely annotated images are used first to train and optimise parameters, and then the optimal parameters are inserted into a Normalized Cut segmentation process. Furthermore, we compare our segmentation results using SNCS to another four classical and two state-of-the-art methods. The overall experimental result shows the usefulness and effectiveness of our method over the six comparison methods.

Vojta-therapy is a useful technique for the treatment of physical and mental impairments in humans, and is very effective for children of less than 6 months. During the therapy, a specific stimulation is given to the patient's body to perform certain reflexive pattern movements. The repetition of this stimulation ultimately makes the previously blocked connections between the spinal cord and brain available, and after a few session, patients can perform these movements without any external stimulation. The treatment must be performed several times a day or week and can last for a few weeks or months. Therefore, the therapists may recommend an at-home continuation of the therapy. An automatic vision-based system is required which can analyze and verify the correct pattern of a patient's body parts movement during the therapy process at home, ultimately revealing the accuracy of given treatment. We captured a dataset of more than 15,000 images from a Microsoft Kinect camera and a novel segmentation technique in RGB-D data is proposed to segment the patient's body region from the scene using a k-means clustering algorithm. The movement patterns of a patient's body parts are analyzed and a support vector machine (SVM) is trained to classify the correct movements. The classification results show that the proposed method is highly useful to recognize the correct movement patterns.

Environmental Microorganisms (EMs) are very diverse and live in every part of the biosphere (rivers, forests, mountains, etc.), playing critical roles in earth's biogeochemical cycles as they are responsible for the decomposition of waste. Currently, a lot of manual efforts through morphological analysis using microscopes have been put on looking for EMs; however, these methods are expensive and time-consuming. To enhance the effectiveness of EM information search, we propose a Content-based Image Retrieval approach by using multiple colour channels fusion and Particle Swarm Optimisation (PSO). First, a microorganism image is decomposed into different colour channels for a more discriminative and eventual representation. Then we extract SIFT features and compute the similarity between a query image and EM database images in terms of each colour channel. Finally, PSO is applied to launch weights fusion and obtain the final retrieval result, because it provides a valid space exploration function. Experiments on our EM dataset show the advantage of the proposed multiple colour channels fusion method over each single channel result.

In this paper, a method for unknown object tracking in output images from 360-degree cameras called Modified Training-Learning-Detection (MTLD) is presented. The proposed method is based on the recently introduced Training-Learning-Detection (TLD) scheme in the literature. The flaws of the TLD approach have been detected and significant modifications are proposed to enhance and to elaborate the scheme. Unlike TLD, MTLD is capable of detecting the unknown objects of interest in 360-degree images. According to the experimental results, the proposed method significantly outperforms the TLD method in terms of detection rate and implementation cost.

Indoor localisation continues to be a topic of growing importance. Despite the advances made, several profound problems are still present. For example, estimating an accurate position from a multimodal distribution or recovering from the influence of faulty measurements. Within this work, we solve such problems with the help of Monte Carlo smoothing methods, namely forward-backward smoother and backward simulation. In contrast to normal filtering procedures like particle filtering, smoothing methods are able to incorporate future measurements instead of just using current and past data. This enables many possibilities for further improving the position estimation. Both smoothing techniques are deployed as fixed-lag and fixed-interval smoother and a novel approach for incorporating them easily within a conventional localisation system is presented. All this is evaluated on four floors within our faculty building. The results show that smoothing methods offer a great tool for improving the overall localisation. Especially fixed-lag smoothing provides a great runtime support by reducing temporal errors and improving the overall estimation with affordable costs.

In this paper, a novel method for polar object tracking in output images from wide field of view of the 360-degree cameras is presented. The proposed method is based on the novel polar object selection to track various objects with diverse shapes directly in polar videos with no rectification. A highly overlapping object selection scheme is proposed for the challenging condition in the moving camera scenarios. Innovative color classifiers are proposed to better detection of desired object in the complex backgrounds. Continuous energy minimizing method is used to deal with the in-plane rotation problem. The selection of a polar area of interest is proposed to optimize the implementation cost. Moreover, the known Lukas-Kanade tracking method is exploited in parallel for further improvement of the tracking result. The experimental results demonstrate the success of the proposed method in terms of precision rate tracking the different objects in the polar images.

Most human detection algorithms in depth images perform well in detecting and tracking the movements of a single human object. However, their performance is rather poor when the person is occluded by other objects or when there are multiple humans present in the scene. In this paper, we propose a novel human detection technique which analyzes the edges in depth image to detect multiple people. The proposed technique detects a human head through a fast template matching algorithm and verifies it through a 3D model fitting technique. The entire human body is extracted from the image by using a simple segmentation scheme comprising a few morphological operators. Our experimental results on three large human detection datasets and the comparison with the state-of-the-art method showed an excellent performance achieving a detection rate of 94.53% with a small false alarm of 0.82%.

We present an indoor localisation system that integrates different sensor modalities, namely Wi-Fi, barometer, iBeacons, step-detection and turn-detection for localisation of pedestrians within buildings over multiple floors. To model the pedestrian's movement, which is constrained by walls and other obstacles, we propose a state transition based upon random walks on graphs. This model also frees us from the burden of frequently updating the system. In addition we make use of barometer information to estimate the current floor. Furthermore, we present a statistical approach to avoid the incorporation of faulty heading information caused by changing the smartphone's position. The evaluation of the system within a 77m × 55m sized building with 4 floors shows that high accuracy can be achieved while also keeping the update-rates low.

Depth estimation is a method to estimate the depth information in a 2D image/video, where the original 3D space is projected onto an image plane. This paper introduces a novel extension of depth estimation in the video domain, where we extract 3D trajectories which individually represent the transition of an object in the 3D space. Such 3D trajectories are useful for appropriately characterising spatio-temporal object relations for video event detection. While we extract 3D trajectories by combining depth estimation and object detection results, the major problem is the inconsistency between these results. For example, significantly different depths may be estimated for the region of the same object, and an object region that is appropriately shaped by estimated depths may be missed. To overcome this, we first initialise the 3D position of an object using the frame with the highest consistency between the depth estimation and object detection results. Then, we track the object in the 3D space using particle filter, where a 3D position of the object is modelled as a hidden state to generate its 2D visual appearance. Experimental results demonstrate the effectiveness of our method.

This paper proposes a novel object matching algorithm based on shape contours. In order to ensure low computational complexity in shape representation, our descriptor is composed by a small number of interest points which are generated by considering both curvatures and the overall shape trend. To effectively describe each point of interest, we introduce a simple and highly discriminative point descriptor, namely Point Context, which represents its geometrical and topological location. For shape matching, we observed that the correspondences are not only dependent on the similarities between these single points in different objects, but they are also related to the geometric relations between multiple points of interest in the same object. Therefore, a high-order graph matching formulation is introduced to merge the single point similarities and the similarities between point triangles. The main contributions of this paper include (i) the introduction of a novel shape descriptor with robust shape points and their descriptors and (ii) the implementation of a high-order graph matching algorithm that solves the shape matching problem. Our method is validated through a series of object retrieval experiments on four datasets demonstrating its robustness and accuracy.

In this paper, we introduce a novel approach to 3D object retrieval by 3D curve matching. First, we project 2D object edges obtained from a depth image into 3D space. Second, we find distinctive feature points on the object. Third, we represent the shortest paths between the features by robust descriptors invariant to rotation, scaling, and translation. Finally, we match two 3D objects using the Maximum Weight Subgraph search. The most important contribution of this paper is the powerful object representation by 3D curves together with the corresponding matching algorithm. Excellent retrieval results achieved with our method show its benefits compared to the state-of-the-art.

In this paper we introduce an approach for object retrieval that uses contour segment matching for shape similarity computation. The object contour is partitioned into segments by skeleton endpoints. Each contour segment is represented by a rotation and scale invariant, 12-dimensional feature vector. The similarity of two objects is determined by matching their contour segments using the Hungarian algorithm. Our method is insensitive to object deformation and outperforms existing shape-based object retrieval algorithms. The most significant scientific contributions of this paper include (i) the introduction of a new feature extraction technique for contour segments as well as (ii) a new similarity measure for contour segments cleverly modelling the human perception and easily adapting to concrete application domains, and (iii) the impressive robustness of the method in an object retrieval scenario.

Occurrence of certain environmental microorganisms and their species is a very informative indicator to evaluate environmental quality. Unfortunately, their manual recognition in microbiological laboratories is very time-consuming and expensive. Therefore, we work on an automatic method for shape-based classification of EMs in microscopic images. First, we segment the microorganisms from the background. Second, we describe their shapes by discriminative feature vectors. Third, we perform the EM classification using Support Vector Machines. The most important scientific contribution of this paper, in comparison to the state-of-the-art and to our previous publications in this field, is the introduction of a completely new and very robust 2D feature descriptor for EM shapes. Experimental results certify the effectiveness and practicability of our automatic EM classification system emphasising the benefits achieved with the new shape descriptor proposed in this work.

Indoor positioning systems are required for many new applications and, ideally, should provide high accuracy at zero costs for initial setup, maintenance and per user. Many approaches thus use an existing Wi-Fi infrastructure and smart-phones for pedestrian location estimation. While the well-known Wi-Fi fingerprinting provides good localization accuracy down to one meter, necessary time and costs are tremendous. Alternatives, like model-based signal strength estimation, are easy to setup but supply viable results only for line of sight conditions. To provide an inexpensive yet accurate solution, we combine Wi-Fi localization using a signal strength prediction model together with step/turn-detection based on a smartphone's accelerometer/gyroscope and incorporate a priori knowledge utilizing the building's floorplan. Our technique uses a statistical model for both, Wi-Fi and step/turn-detection, to calculate the probability of the pedestrian residing at some arbitrary position and leverages the building's floorplan to determine the likelihood for any possible movement between two positions. Latter probabilities are combined with the two densities from Wi-Fi and step/turn-detection applying recursive density estimation implemented using well-known particle filtering techniques. While the density estimated through Wi-Fi measurements provides a vague, absolute position at a significant uncertainty, the step-detector supplies a fine resolution at the downside of a cumulative error due to its estimation relative to previous steps. The fusion of these two sensor densities compensates for this error and the floorplan further enhances the estimation result. We will show that our statistical approach provides a robust, long-term stable location estimation, much better than Wi-Fi on its own while requiring only a few (empirical) parameters.

This paper presents a method for indoor localization of humans. Our new approach combines an imaged-based position estimation with a given step and turn detection. Estimating the position of an object is not only a question of accuracy, it is also a question of performance and time required. Our approach uses a fusion of an uncalibrated depth-sensor with a smartphone's accelerometer and gyroscope. Both components do not rely on time-consuming practices like fingerprinting and calibration techniques. This approach allows for a real time-tracking by using well-known methods of recursive density propagation and particle filtering. Unlike other image-based methods in autonomous robotics, the depth sensors are mounted at a fixed position. Therefore we will show how our new statistical sensor model covers the four main conditions of an image-based approach: a person is either inside or outside the field of view and is detected or not detected by the depth sensor. This involves the possibility to estimate the position of multiple persons at each point in time. Finally, the experimental results show how the integration of an image-based approach increases the accuracy of the localization estimation and counteracts the increasing error of the given step and turn detection.

This paper introduces a method for Multimedia Event Detection (MED). Given training videos for a certain event, a classifier is constructed to identify videos displaying it. In particular, the problems of the weakly supervised setting and the unclear event structure are addressed in this paper. The first issue is associated with the loosely annotated training videos that usually contain many irrelevant shots. The second one is the difficulty of assuming the event structure in advance, because videos can be created by arbitrary camera and editing techniques. To overcome these problems, a Hidden Conditional Random Field (HCRF) is used where hidden states work as an intermediate layer to discriminate between relevant and irrelevant shots to the event. In addition, the relation among hidden states characterises the event structure. Thus, the above problems are managed by optimising hidden states and their relation, so as to distinguish videos where the event occurs from the rest of videos. Experimental results on TRECVID video data validate the effectiveness of HCRFs in this context.

Environmental Microorganisms (EMs) are currently recognised using molecular biology (DNA, RNA) or morphological methods. The first ones are very time-consuming and expensive. The second ones require a very experienced laboratory operator. To overcome these problems, we introduce an automatic classification method for EMs in the framework of content-based image analysis in this paper. To describe the shapes of EMs observed in microscopic images, we use Edge Histograms, Fourier Descriptors, extended Geometrical Features, as well as introduce Internal Structure Histograms. For classification, multi-class Support Vector Machine is applied to EMs represented by the above features. In order to quantitatively evaluate discriminative properties of the feature spaces we have introduced, we perform comprehensive experiments with a ground truth of manually segmented microscopic EM images. The best classification result of 89.7% proves a high robustness of our method in this application domain.

The objective of this special issue is to report on recent trends in digital and media technologies responding to the challenges of managing and accessing multimedia (images, audio, video, 3D/4D material, etc.).In a highly selective review procedure we accepted contributions describing recent work that aims at narrowing the large disparity between the low-level multimedia descriptors and the richness of subjectivity of semantics in user queries and human interpretations of audiovisual data.The articles in this special issue can be grouped into following categories: Multimedia Analysis (Section 1), Multimedia Ontologies and Data Integration (Section 2), and Social Media and Retrieval (Section 3).

A virtual articulator is a system currently being developed for dentistry. It is supposed to virtually determine teeth antagonists and the occlusion surfaces between them. In contrast to state-of-the-art mechanical articulators, it should use 3D dentition models acquired with highly precise 3D scanners for this. These models, however, describe not only teeth surfaces. Neighbouring surfaces such as teethridge are also included in them. Thus, the first and unavoidable step in designing a virtual articulator is the segmentation of teeth from 3D dentition models. In this paper, we present a robust solution to this problem. First, 2D sectional images of the 3D models are generated and binarised. Second, 2D teeth outlines are estimated using active contours and summarised to 3D segmentation results. The algorithm has provided excellent experimental results and will be used in the industrial realisation of the virtual articulator.