International Journal of Pattern Recognition and Artificial Intelligence

Hierarchical image segmentation provides a region-oriented scale-space, i.e. a set of image segmentations at different detail levels in which the segmentations at finer levels are nested with respect to those at coarser levels. However, most image segmentation algorithms, among which a graph-based image segmentation method relying on a region merging criterion was proposed by Felzenszwalb–Huttenlocher in 2004, do not lead to a hierarchy. In order to cope with a demand for hierarchical segmentation, Guimarães et al. proposed in 2012 a method for hierarchizing the popular Felzenszwalb–Huttenlocher method, without providing an algorithm to compute the proposed hierarchy. This paper is devoted to providing a series of algorithms to compute the result of this hierarchical graph-based image segmentation method efficiently, based mainly on two ideas: optimal dissimilarity measuring and incremental update of the hierarchical structure. Experiments show that, for an image of size 321 × 481 pixels, the most efficient algorithm produces the result in half a second whereas the most naive one requires more than 4 h.

Class imbalance refers to classification problems in which many more instances are available for certain classes than for others. Such imbalanced datasets require special attention because traditional classifiers generally favor the majority class which has a large number of instances. Ensemble of classifiers has been reported to yield promising results. However, the majority of ensemble methods applied to imbalance learning are static ones. Moreover, they only deal with binary imbalanced problems. Hence, this paper presents an empirical analysis of Dynamic Selection techniques and data preprocessing methods for dealing with multi-class imbalanced problems. We considered five variations of preprocessing methods and 14 Dynamic Selection schemes. Our experiments conducted on 26 multi-class imbalanced problems show that the dynamic ensemble improves the AUC and the $G$-mean as compared to the static ensemble. Moreover, data preprocessing plays an important role in such cases.

Human visual perception has a close relationship with the HSV color space, which can be represented as a cylinder. The question of how visual features are extracted using such an attribute is important. In this paper, a new feature descriptor; namely, a color volume histogram, is proposed for image representation and content-based image retrieval. It converts a color image from RGB color space to HSV color space and then uniformly quantizes it into 72 bins of color cues and 32 bins of edge cues. Finally, color volumes are used to represent the image content. The proposed algorithm is extensively tested on two Corel datasets containing 15000 natural images. These image retrieval experiments show that the color volume histogram has the power to describe color, texture, shape and spatial features and performs significantly better than the local binary pattern histogram and multi-texton histogram approaches.

In this paper, we consider a mobile platform with two cameras directed towards the floor. In earlier work, this specific problem geometry has been considered under the assumption that the cameras have been mounted at the same height. This paper extends the previous work by removing the height constraint, as it is hard to realize in real-life applications.

We develop a method based on an equivalent problem geometry, and show that much of previous work can be reused with small modification to account for the height difference. A fast solver for the resulting nonconvex optimization problem is devised. Furthermore, we propose a second method for estimating the height difference by constraining the mobile platform to pure translations. This is intended to simulate a calibration sequence, which is not uncommon to impose. Experiments are conducted using synthetic data, and the results demonstrate a robust method for determining the relative parameters comparable to previous work.

The standard representations known as component trees, used in morphological connected attribute filtering and multi-scale analysis, are unsuitable for cases in which either the image itself or the tree do not fit in the memory of a single compute node. Recently, a new structure has been developed which consists of a collection of modified component trees, one for each image tile. It has to-date only been applied to fairly simple image filtering based on area. In this paper, we explore other applications of these distributed component forests, in particular to multi-scale analysis such as pattern spectra, and morphological attribute profiles and multi-scale leveling segmentations.

Early and accurate Alzheimer’s disease (AD) diagnosis remains a challenge. Recently, increasing efforts have been focused towards utilization of metabolomics data for the discovery of biomarkers for screening and diagnosis of AD. Several machine learning approaches were explored for classifying the blood metabolomics profiles of cognitively healthy and AD patients. Differentiation between AD, mild cognitive impairment (MCI) and cognitively healthy subjects remains difficult. In this paper, we propose a new machine learning approach for the selection of a subset of features that provide an improvement in classification rates between these three levels of cognitive disorders. Our experimental results demonstrate that utilization of these selected metabolic markers improves the performance of several classifiers in comparison to the classification accuracy obtained for the complete metabolomics dataset. The obtained results indicate that our algorithms are effective in discovering a panel of biomarkers of AD and MCI from metabolomics data suggesting the possibility to develop a noninvasive blood diagnostic technique of AD and MCI.

Thousands of criminal events are reported in newspapers and social networks every day. They describe violent acts that include actors, places, times, causes and any information concerning them. Verbal and nominal phrases are used to characterize and expose criminal events, which employ an important variety of natural language structures in the newspapers. In addition, causes, times and spaces of criminal events, use linguistic phrases to represent them in text. All of them need to be extracted as a pattern recognition process in order to extract criminal events from text and the information that concerns them. The extracted events, as a knowledge base, are very useful for information retrieval tasks. Therefore, this paper presents an approach based on pattern recognition in order to extract criminal events from Spanish text, by populating and enriching an ontology model. Ontology population and enrichment involve the instantiation of criminal events and their cause relationships. An evaluation process is carried out with a set of manually tagged newspapers with categories of specific events, and shows promising results.

Emotion recognition plays an indispensable role in human–machine interaction system. The process includes finding interesting facial regions in images and classifying them into one of seven classes: angry, disgust, fear, happy, neutral, sad, and surprise. Although many breakthroughs have been made in image classification, especially in facial expression recognition, this research area is still challenging in terms of wild sampling environment. In this paper, we used multi-level features in a convolutional neural network for facial expression recognition. Based on our observations, we introduced various network connections to improve the classification task. By combining the proposed network connections, our method achieved competitive results compared to state-of-the-art methods on the FER2013 dataset.

With the outstanding performance in 2014 at the ImageNet Large-Scale Visual Recognition Challenge 2014 (ILSVRC14), an effective convolutional neural network (CNN) model named GoogLeNet has drawn the attention of the mainstream machine learning field. In this paper we plan to take an insight into the application of the GoogLeNet in the Handwritten Chinese Character Recognition (HCCR) on the database HCL2000 and CASIA-HWDB with several necessary adjustments and also state-of-the-art improvement methods for this end-to-end approach. Through the experiments we have found that the application of the GoogLeNet for the Handwritten Chinese Character Recognition (HCCR) results into significant high accuracy, to be specific more than 99% for the final version, which is encouraging for us to further research.

Deep residual network (ResNet) is currently the basis of many popular state-of-the-art convolutional neural network models for image recognition, and its recent variants include wide residual network (WRN), aggregated deep residual network (ResNeXt) and deep pyramidal residual network (PyramidNet). Here, we demonstrate the potential application of deep residual network and its variants in high-content screening (i.e. cell phenotype classification) that can overcome issues associated with analyzing high-content screening data, such as exhaustive preprocessing and inefficient learning. Cell phenotype classification is an image-based method that can be used for drug high-content screening, in which complex cell states associated with chemical compound treatment can be characterized. Previous work on cell phenotype classification typically requires a routine yet cumbersome step of single cell segmentation before the classification task. In this paper, we present a segmentation-free method for image-based cell phenotype classification using deep ResNet and its variants. The cell images are samples treated with annotated compounds that can be mainly grouped into three clusters, giving three classes to be classified. Instead of single-cell phenotype classification, we use the raw images without segmentation for our training and evaluation directly. Compared to previous reference work, we significantly simplify the data preprocessing step and accelerate the training while still achieving high accuracy. Our trained models achieve a 98.2% accuracy rate on the three classes classification problem (three compound clusters only), and a 93.8% accuracy rate on the four classes classification problem (three compound clusters plus the mock class) based on five-fold cross-validation.

The paper is devoted to Descriptive Image Analysis (DA) — a leading line of the modern mathematical theory of image analysis. DA is a logically organized set of descriptive methods, mathematical objects, and models and representations aimed at analyzing and evaluating the information represented in the form of images, as well as for automating the extraction from images of knowledge and data needed for intelligent decision-making.

The basic idea of DA consists of embedding all processes of analysis (processing, recognition, understanding) of images into an image formalization space and reducing it to (1) construction of models/representations/formalized descriptions of images; (2) construction of models/representations/formalized descriptions of transformations over models and representations of images.

We briefly discuss the basic ideas, methodological principles, mathematical methods, objects, and components of DA and the basic results determining the current state of the art in the field. Image algebras (IA) are considered in the context of a unified language for describing mathematical objects and operations used in image analysis (the standard IA by Ritter and the descriptive IA by Gurevich).

In this work, the sensor data related to a diesel engine system and specifically its turbocharger subsystem were analyzed. An incident where the turbocharger seized was recorded by dozens of standard turbocharger-related sensors. By training models to distinguish between normal healthy operating conditions and deteriorated conditions, there is an opportunity to develop prognostic and predictive tools to ideally help prevent a similar occurrence in the future. Analysis of this event provides an opportunity to identify changes in equipment indicators with a known outcome. A number of data analysis tools were used to characterize the healthy and deteriorated states of the turbocharger system, including various supervised classification as well as semi-supervised and unsupervised anomaly detection techniques. The leader clustering algorithm was also implemented to reduce the amount of data to train and develop the models. This paper describes the results of this modeling process, validated by testing on healthy data from the same propulsion system and a second distinct one. Although this problem posed challenges due to the severely imbalanced class distribution, the supervised classifiers, in particular Support Vector Machine (SVM) and Random Forest (RF), performed very well in all metrics while the unsupervised anomaly detection models achieved near-perfect accuracy for identifying healthy turbocharger states.

In bioinformatics, the biological functions of proteins and their interactions can often be analyzed by the similarity of their sequences. In this paper, the authors combine the fractal dimension, empirical mode decomposition (EMD), and sliding window for protein sequence comparison. First, the protein sequence is characterized and digitized into a signal, and then the signal characteristics are obtained by using EMD and fractal dimension. Each protein sequence can be decomposed into Intrinsic Mode Functions (IMFs). The fixed window’s fractal dimension is applied to each IMF and the original signal to extract the protein sequence characteristics. Experiments have shown that the feature extracted by this hybrid method is superior to the EMD method alone.

In firearm identification, a firearm examiner looks at a pair of fired bullets or cartridge cases using a comparison microscope and determines from this visual analysis if they were both fired from the same firearm. In the particular case of fired bullets, the individual firearm signature takes the form of a striated pattern. Over the time, the firearm examiner’s community developed two distinct approaches for bullet identification: pattern matching and line counting. More recently, the emergence of technology enabling the capture of surface topographies down to a submicron depth resolution has been a catalyst for the field of computerized objective ballistic identification. Objectiveness is achieved through the statistical analysis of various scores of known matches and known nonmatches exhibit pair comparison, which in turn implies the capture of large quantities of bullets and cartridge cases topographies. The main goal of this study was to develop an objective identification method for bullets fired from conventionally rifled barrels, and to test this method on public and proprietary bullet 3D image datasets captured at different lateral resolutions. Two newly developed bullet identification scores, the Line Counting Score (LCS) and the Pattern Matching Score, computed on 3D topographies yielded perfect match versus nonmatch separation for three different sets used in the standard Hamby–Brundage Test. A similar analysis performed using a larger, more-realistic set, enabled us to define a discriminative line at a false match rate of 1/10000 on a 2D plot that shows both identification scores for matches and nonmatches. The LCS is shown to produce a better sensitivity than the standard consecutive matching striae criteria for the more-realistic dataset. A likelihood function was also computed from a linear combination of both scores, and a conservative approach based on extreme value theory is proposed to extrapolate this function in the score domain where nonmatch data are not available. This study also provides a better understanding of the limitations of studies that involve very few firearms.

Finger-vein verification is a highly secure biometric authentication that has been widely investigated over the last years. One of its challenges, however, is the possible degradation of image quality, that results in spurious and missing vein patterns, which increases the verification error. Despite recent advances in finger-vein quality assessment, the proposed solutions are limited as they depend on human expertise and domain knowledge to extract handcrafted features for assessing quality. We have proposed, recently, the first deep neural network (DNN) framework for assessing finger-vein quality, that does not require manual labeling of high and low quality images, as is the case for state of the art methods, but infers such annotations automatically based on an objective indicator, the biometric verification decision. This framework has significantly outperformed the existing methods, whether the input image is in grayscale or is binary. Motivated by these performances, we propose, in this work, a representation learning of finger vein image quality, where a DNN takes as input conjointly the grayscale and binary versions of the input image to predict vein quality. Our model allows to learn the joint representation from grayscale and binary images, for quality assessment. The experimental results, obtained on a large public dataset, demonstrates that our proposed method accurately identifies high and low quality images, and outperforms other techniques in terms of equal error rate (EER) minimization, including our previous DNN models, based either on grayscale or binary input.