Deep Neural Networks (DNNs) often struggle to generalize beyond their training distributions, making them vulnerable to domain shifts. To enhance robustness, various approaches have been developed, particularly focusing on data-centric methods. Modifying the training data can increase diversity and improve generalization, while also introducing bias that positively guides the model’s decision-making. Prior research suggests that DNNs tend to overemphasize texture-based patterns, at the expense of more robust shape-based representations. We introduce ShapeBlend, a novel data augmentation technique that emphasizes image contours, and hence shape features. It blends a contour map from the push-pull CORF operator with the original image at varying strengths. ShapeBlend consistently outperforms state-of-the-art methods across major Out-of-Distribution (OOD) benchmarks (ImageNet-A, ImageNet-R, ImageNet-C, and ImageNet- C ¯ ), setting new records in robustness. Moreover, ShapeBlend’s versatility allows its application during inference. To fully leverage ShapeBlend, we propose Shape-Enhanced Voting (SEV), an inference strategy that aggregates predictions from multiple ShapeBlend-processed images. The combination of ShapeBlend and SEV further enhances domain robustness, with performance gains varying based on the chosen configuration. • We introduce ShapeBlend: a shape-based augmentation for OOD robustness. • ShapeBlend improves performance across multiple robustness benchmarks. • Inference performance is boosted with Shape-enhanced Voting (SEV). • ShapeBlend is compatible with existing pipelines like AugMix and DeepAugment. • A simple, explainable method with strong theoretical motivation.

Pedestrian attribute recognition has become an important task in computer vision, particularly for retail and marketing and critical applications in security and surveillance. Despite its potential, achieving real-time performance on embedded devices while maintaining accuracy has been a significant challenge. In this paper, we propose a novel multi-task method for simultaneous person attribute recognition using task-specific attention network, which shares low-level representations across related tasks, reducing computational and memory requirements without compromising accuracy. In particular, we employ a spatial-channel attention mechanism to selectively focus on relevant regions without increasing the computational complexity of the backbone. Furthermore, we use a knowledge distillation technique to deal with missing labels and gradient normalization for dealing with task imbalances, since varying task difficulties lead to disproportionate gradient magnitudes during training. The experimental results demonstrate the effectiveness of our approach, achieving a mean accuracy of 0.889 while maintaining real-time performance at 114 frames per second on an embedded board with limited resources. These results highlight the practical viability and novelty of our system as a robust and scalable solution for pedestrian attribute recognition on embedded devices in real-world scenarios.

Modern computer vision technologies enable systems to detect, recognize, and analyze facial features, but challenges arise when images are noisy, blurred, or low quality. Blind face restoration, which aims to recover high-quality facial images without prior knowledge of degradation, addresses this issue. In this paper, we introduce Fair Restoration GAN (FaiResGAN), a novel Generative Adversarial Network (GAN) designed to balance face restoration with the preservation of soft biometrics (identity, ethnicity, age, and gender). Our model incorporates a pseudo-random batch composition algorithm to promote fairness and mitigate bias, alongside a realistic degradation model simulating corruptions typical in surveillance images. Experimental results show that FaiResGAN outperforms state-of-the-art blind face restoration methods, both quantitatively and qualitatively. A user study involving 40 participants showed that FaiResGAN-restored images were preferred by 70% of users. Additionally, tests on VGGFace2, UTKFace, and FairFace datasets demonstrate FaiResGAN’s superior performance in preserving soft biometric attributes and ensuring fair restoration across different genders and ethnicities.

Abstract Ensemble learning for artificial neural networks (ANNs) is an effective method to enhance predictive performance. However, ANNs are computationally and memory intensive, and naively training multiple networks can lead to excessive training times and costs. An effective tool for improving ensemble efficiency is introducing topological sparsity. Even though several implementations of efficient ensembles have been proposed, none of them can provide actual benefits in terms of computational overhead as the sparsity is simulated using binary masks. In this paper, we address this issue by introducing a Truly Sparse Ensemble without binary masks and directly incorporate native sparsity. We also propose two algorithms for initializing new subnetworks within the ensemble, leveraging this native topological sparsity to enhance subnetwork diversity. We demonstrate the performance of the resulting models at high levels of sparsity on several datasets in terms of classification accuracy, floating point operations (FLOPs), and actual running time. The proposed methods outperform all baseline dense and truly sparse models on tabular data, successfully diversify the training trajectory of the subnetworks, and increase the topological distance between subnetworks after re-initialization.

Abstract We propose a novel and versatile computational approach, based on hierarchical COSFIRE filters, that addresses the challenge of explainable retina and palmprint recognition for automatic person identification. Unlike traditional systems that treat these biometrics separately, our method offers a unified solution, leveraging COSFIRE filters’ trainable nature for enhanced selectivity and robustness, while exhibiting explainability and resilience to decision-based black-box adversarial attack and partial matching. COSFIRE filters are trainable, in that their selectivity can be determined with a one-shot learning step. In practice, we configure a COSFIRE filter that is selective for the mutual spatial arrangement of a set of automatically selected keypoints of each retina or palmprint reference image. A query image is then processed by all COSFIRE filters and it is classified with the reference image that was used to configure the COSFIRE filter that gives the strongest similarity score. Our approach, tested on the VARIA and RIDB retina datasets and the IITD palmprint dataset, achieved state-of-the-art results, including perfect classification for retina datasets and a 97.54% accuracy for the palmprint dataset. It proved robust in partial matching tests, achieving over 94% accuracy with 80% image visibility and over 97% with 90% visibility, demonstrating effectiveness with incomplete biometric data. Furthermore, while effectively resisting a decision-based black-box adversarial attack and impervious to imperceptible adversarial images, it is only susceptible to highly perceptible adversarial images with severe noise, which pose minimal concern as they can be easily detected through histogram analysis in preprocessing. In principle, the proposed learning-free hierarchical COSFIRE filters are applicable to any application that requires the identification of certain spatial arrangements of moderately complex features, such as bifurcations and crossovers. Moreover, the selectivity of COSFIRE filters is highly intuitive; and therefore, they provide an explainable solution.

Directly mining high-dimensional time series presents several challenges, such as time and space costs. This study proposes a new approach for representing time series data and evaluates its effectiveness in detecting collective anomalies. The proposed method, called bidirectional piecewise linear representation (BPLR), represents the original time series using a set of linear fitting functions, which allows for dimensionality reduction while maintaining its dynamic characteristics. Similarity measurement is then performed using the piecewise integration (PI) approach, which achieves good detection performance with low computational overhead. Experimental results on synthetic and real-world data sets confirm the effectiveness and advantages of the proposed approach. The ability of the proposed method to capture more dynamic details of time series leads to consistently superior performance compared to other existing methods.

Electron microscopy (EM) enables high-resolution imaging of tissues and cells based on 2D and 3D imaging techniques. Due to the laborious and time-consuming nature of manual segmentation of large-scale EM datasets, automated segmentation approaches are crucial. This review focuses on the progress of deep learning-based segmentation techniques in large-scale cellular EM throughout the last six years, during which significant progress has been made in both semantic and instance segmentation. A detailed account is given for the key datasets that contributed to the proliferation of deep learning in 2D and 3D EM segmentation. The review covers supervised, unsupervised, and self-supervised learning methods and examines how these algorithms were adapted to the task of segmenting cellular and sub-cellular structures in EM images. The special challenges posed by such images, like heterogeneity and spatial complexity, and the network architectures that overcame some of them are described. Moreover, an overview of the evaluation measures used to benchmark EM datasets in various segmentation tasks is provided. Finally, an outlook of current trends and future prospects of EM segmentation is given, especially with large-scale models and unlabeled images to learn generic features across EM datasets.

As dimensions of datasets in predictive modelling continue to grow, feature selection becomes increasingly practical. Datasets with complex feature interactions and high levels of redundancy still present a challenge to existing feature selection methods. We propose a novel framework for feature selection that relies on boosting, or sample re-weighting, to select sets of informative features in classification problems. The method uses as its basis the feature rankings derived from fast and scalable tree-boosting models, such as XGBoost. We compare the proposed method to standard feature selection algorithms on 9 benchmark datasets. We show that the proposed approach reaches higher accuracies with fewer features on most of the tested datasets, and that the selected features have lower redundancy.

A crucial challenge in digital image forensics is to identify the source camera model used to generate given images. This is of prime importance, especially for Law Enforcement Agencies in their investigations of Child Sexual Abuse Material found in darknets or seized storage devices. In this work, we address this challenge by proposing a solution that is characterized by two main contributions. It relies on the extraction of rather small homogeneous regions that we extract very efficiently from the integral image, and on a hierarchical classification approach with convolutional neural networks as the underlying models. We rely on homogeneous regions as they contain camera traces that are less distorted than regions with high-level scene content. The hierarchical approach that we propose is important for scaling up and making minimal modifications when new cameras are added. Furthermore, this scheme performs better than the traditional single classifier approach. By means of thorough experimentation on the publicly available Dresden data set, we achieve an accuracy of 99.01% with 5-fold cross-validation on the ‘natural’ subset of this data set. To the best of our knowledge, this is the best result ever reported for Dresden data set.

Livestock management involves the monitoring of farm animals by tracking certain physiological and phenotypical characteristics over time. In the dairy industry, for instance, cattle are typically equipped with RFID ear tags. The corresponding data (e.g. milk properties) can then be automatically assigned to the respective cow when they enter the milking station. In order to move towards a more scalable, affordable, and welfare-friendly approach, automatic non-invasive solutions are more desirable. Thus, a non-invasive approach is proposed in this paper for the automatic identification of individual Holstein cattle from the side view while exiting a milking station. It considers input images from a thermal-RGB camera. The thermal images are used to delineate the cow from the background. Subsequently, any occluding rods from the milking station are removed and inpainted with the fast marching algorithm. Then, it extracts the RGB map of the segmented cattle along with a novel CORF3D contour map. The latter contains three contour maps extracted by the Combination of Receptive Fields (CORF) model with different strengths of push–pull inhibition. This mechanism suppresses noise in the form of grain type texture. The effectiveness of the proposed approach is demonstrated by means of experiments using a 5-fold and a leave-one day-out cross-validation on a new data set of 3694 images of 383 cows collected from the Dairy Campus in Leeuwarden (the Netherlands) over 9 days. In particular, when combining RGB and CORF3D maps by late fusion, an average accuracy of 99.64%(±0.13) was obtained for the 5-fold cross validation and 99.71%(±0.31) for the leave-one day-out experiment. The two maps were combined by first learning two ConvNet classification models, one for each type of map. The feature vectors in the two FC layers obtained from training images were then concatenated and used to learn a linear SVM classification model. In principle, the proposed approach with the novel CORF3D contour maps is suitable for various image classification applications, especially where grain type texture is a confounding variable.

The identification of source cameras from videos, though it is a highly relevant forensic analysis topic, has been studied much less than its counterpart that uses images. In this work we propose a method to identify the source camera of a video based on camera specific noise patterns that we extract from video frames. For the extraction of noise pattern features, we propose an extended version of a constrained convolutional layer capable of processing color inputs. Our system is designed to classify individual video frames which are in turn combined by a majority vote to identify the source camera. We evaluated this approach on the benchmark VISION data set consisting of 1539 videos from 28 different cameras. To the best of our knowledge, this is the first work that addresses the challenge of video camera identification on a device level. The experiments show that our approach is very promising, achieving up to 93.1% accuracy while being robust to the WhatsApp and YouTube compression techniques. This work is part of the EU-funded project 4NSEEK focused on forensics against child sexual abuse.

One of the challenging problems in digital image forensics is the capability to identify images that are captured by the same camera device. This knowledge can help forensic experts in gathering intelligence about suspects by analyzing digital images. In this paper, we propose a two-part network to quantify the likelihood that a given pair of images have the same source camera, and we evaluated it on the benchmark Dresden data set containing 1851 images from 31 different cameras. To the best of our knowledge, we are the first ones addressing the challenge of device-based image matching. Though the proposed approach is not yet forensics ready, our experiments show that this direction is worth pursuing, achieving at this moment 85 percent accuracy. This ongoing work is part of the EU-funded project 4NSEEK concerned with forensics against child sexual abuse.

The recent technological advent of cryptocurrencies and their respective benefits have been shrouded with a number of illegal activities operating over the network such as money laundering, bribery, phishing, fraud, among others. In this work we focus on the Ethereum network, which has seen over 400 million transactions since its inception. Using 2179 accounts flagged by the Ethereum community for their illegal activity coupled with 2502 normal accounts, we seek to detect illicit accounts based on their transaction history using the XGBoost classifier. Using 10 fold cross-validation, XGBoost achieved an average accuracy of 0.963 ( ± 0.006) with an average AUC of 0.994 ( ± 0.0007). The top three features with the largest impact on the final model output were established to be ‘Time diff between first and last (Mins)’, ‘Total Ether balance’ and ‘Min value received’. Based on the results we conclude that the proposed approach is highly effective in detecting illicit accounts over the Ethereum network. Our contribution is multi-faceted; firstly, we propose an effective method to detect illicit accounts over the Ethereum network; secondly, we provide insights about the most important features; and thirdly, we publish the compiled data set as a benchmark for future related works.

Contour detection is a salient operation in many computer vision applications as it extracts features that are important for distinguishing objects in scenes. It is believed to be a primary role of simple cells in visual cortex of the mammalian brain. Many of such cells receive push-pull inhibition or surround suppression. We propose a computational model that exhibits a combination of these two phenomena. It is based on two existing models, which have been proven to be very effective for contour detection. In particular, we introduce a brain-inspired contour operator that combines push-pull and surround inhibition. It turns out that this combination results in a more effective contour detector, which suppresses texture while keeping the strongest responses to lines and edges, when compared to existing models. The proposed model consists of a Combination of Receptive Field (or CORF) model with push-pull inhibition, extended with surround suppression. We demonstrate the effectiveness of the proposed approach on the RuG and Berkeley benchmark data sets of 40 and 500 images, respectively. The proposed push-pull CORF operator with surround suppression outperforms the one without suppression with high statistical significance.

Delineation of curvilinear structures in images is an important basic step of several image processing applications, such as segmentation of roads or rivers in aerial images, vessels or staining membranes in medical images, and cracks in pavements and roads, among others. Existing methods suffer from insufficient robustness to noise. In this paper, we propose a novel operator for the detection of curvilinear structures in images, which we demonstrate to be robust to various types of noise and effective in several applications. We call it RUSTICO, which stands for RobUST Inhibition-augmented Curvilinear Operator. It is inspired by the push-pull inhibition in visual cortex and takes as input the responses of two trainable B-COSFIRE filters of opposite polarity. The output of RUSTICO consists of a magnitude map and an orientation map. We carried out experiments on a data set of synthetic stimuli with noise drawn from different distributions, as well as on several benchmark data sets of retinal fundus images, crack pavements, and aerial images and a new data set of rose bushes used for automatic gardening. We evaluated the performance of RUSTICO by a metric that considers the structural properties of line networks (connectivity, area, and length) and demonstrated that RUSTICO outperforms many existing methods with high statistical significance. RUSTICO exhibits high robustness to noise and texture.

Musical notation is a means of passing on performance instructions with fidelity to others. Composers, however, often introduced embellishments to the music they performed notating these embellishments with symbols next to the relevant notes. In time, these symbols, known as ornaments, and their interpretation became standardized such that there are acceptable ways of interpreting an ornament. Although music books may contain footnotes which express the ornament in full notation, these remain cumbersome to read. Ideally, a music student will have the possibility of selecting ornamented notes and express them as full notation. The student should also have the possibility to collapse the expressed ornament back to its symbolic representation, giving the student the possibility of also becoming familiar with playing from the ornamented score. In this paper, we propose a complete pipeline that achieves this goal. We compare the use of COSFIRE and template matching for optical music recognition to identify and extract musical content from the score. We then express the score using MusicXML and design a simple user interface which allows the user to select ornamented notes, view their expressed notation and decide whether they want to retain the expressed notation, modify it, or revert to the symbolic representation of the ornament. The performance results that we achieve indicate the effectiveness of our proposed approach.

Engineering design makes use of freehand sketches to communicate ideas, allowing designers to externalise form concepts quickly and naturally. Such sketches serve as working documents which demonstrate the evolution of the design process. For the product design to progress, however, these sketches are often redrawn using computer-aided design tools to obtain virtual, interactive prototypes of the design. Although there are commercial software packages which extract the required information from freehand sketches, such packages typically do not handle the complexity of the sketched drawings, particularly when considering the visual cues that are introduced to the sketch to aid the human observer to interpret the sketch. In this paper, we tackle one such complexity, namely the use of shading and shadows which help portray spatial and depth information in the sketch. For this reason, we propose a vectorisation algorithm, based on trainable COSFIRE filters for the detection of junction points and subsequent tracing of line paths to create a topology graph as a representation of the sketched object form. The vectorisation algorithm is evaluated on 17 sketches containing different shading patterns and drawn by different sketchers specifically for this work. Using these sketches, we show that the vectorisation algorithm can handle drawings with straight or curved contours containing shadow cues, reducing the salient point error in the junction point location by 91% of that obtained by the off-the-shelf Harris-Stephen's corner detector while the overall vectorial representations of the sketch achieved an average F-score of 0.92 in comparison to the ground truth. The results demonstrate the effectiveness of the proposed approach.

Gender recognition from face images is an important application and it is still an open computer vision problem, even though it is something trivial from the human visual system. Variations in pose, lighting, and expression are few of the problems that make such an application challenging for a computer system. Neurophysiological studies demonstrate that the human brain is able to distinguish men and women also in absence of external cues, by analyzing the shape of specific parts of the face. In this paper, we describe an automatic procedure that combines trainable shape and color features for gender classification. In particular the proposed method fuses edge-based and color-blob-based features by means of trainable COSFIRE filters. The former types of feature are able to extract information about the shape of a face whereas the latter extract information about shades of colors in different parts of the face. We use these two sets of features to create a stacked classification SVM model and demonstrate its effectiveness on the GENDER-COLOR-FERET dataset, where we achieve an accuracy of 96.4%.

The growing interest in recent years for gender recognition from face images is mainly attributable to the wide range of possible applications that can be used for commercial and marketing purposes. It is desirable that such algorithms process high resolution video frames acquired by using surveillance cameras in real-time. To the best of our knowledge, however, there are no studies which analyze the computational impact of the methods and the difficulties related to the processing of faces extracted from videos captured in the wild. We propose a novel face descriptor based on the gradient magnitudes of facial landmarks, which are points automatically extracted from the face contour, eyes, eyebrows, nose, mouth and chin. We evaluate the effectiveness and efficiency of the proposed approach on two new datasets, which we made available online and that consist of color face images and color video sequences acquired in real scenarios. The proposed approach is more efficient and effective than three commercial libraries.

Most object recognition methods rely on contour-defined features obtained by edge detection or region segmentation. They are not robust to diffuse region boundaries. Furthermore, such methods do not exploit region color information. We propose color-blob-based COSFIRE (Combination of Shifted Filter Responses) filters to be selective for combinations of diffuse circular regions (blobs) in specific mutual spatial arrangements. Such a filter combines the responses of a certain selection of Difference-of-Gaussians filters, essentially blob detectors, of different scales, in certain channels of a color space, and at certain relative positions to each other. Its parameters are determined/learned in an automatic configuration process that analyzes the properties of a given prototype object of interest. We use these filters to compute features that are effective for the recognition of the prototype objects. We form feature vectors that we use with an SVM classifier. We evaluate the proposed method on a traffic sign (GTSRB) and a butterfly data sets. For the GTSRB data set we achieve a recognition rate of 98.94%, which is slightly higher than human performance and for the butterfly data set we achieve 89.02%. The proposed color-blob-based COSFIRE filters are very effective and outperform the contour-based COSFIRE filters. A COSFIRE filter is trainable, it can be configured with a single prototype pattern and it does not require domain knowledge.

Gender recognition from face images is an important application in the fields of security, retail advertising and marketing. We propose a novel descriptor based on COSFIRE filters for gender recognition. A COSFIRE filter is trainable, in that its selectivity is determined in an automatic configuration process that analyses a given prototype pattern of interest. We demonstrate the effectiveness of the proposed approach on a new dataset called GENDER-FERET with 474 training and 472 test samples and achieve an accuracy rate of 93.7%. It also outperforms an approach that relies on handcrafted features and an ensemble of classifiers. Furthermore, we perform another experiment by using the images of the Labeled Faces in the Wild (LFW) dataset to train our classifier and the test images of the GENDER-FERET dataset for evaluation. This experiment demonstrates the generalization ability of the proposed approach and it also outperforms two commercial libraries, namely Face++ and Luxand.

The recently proposed trainable COSFIRE filters are highly effective in a wide range of computer vision applications, including object recognition, image classification, contour detection and retinal vessel segmentation. A COSFIRE filter is selective for a collection of contour parts in a certain spatial arrangement. These contour parts and their spatial arrangement are determined in an automatic configuration procedure from a single user-specified pattern of interest. The traditional configuration, however, does not guarantee the selection of the most distinctive contour parts. We propose a genetic algorithm-based optimization step in the configuration of COSFIRE filters that determines the minimum subset of contour parts that best characterize the pattern of interest. We use a public dataset of images of an edge milling head machine equipped with multiple cutting tools to demonstrate the effectiveness of the proposed optimization step for the detection and localization of such tools. The optimization process that we propose yields COSFIRE filters with substantially higher generalization capability. With an average of only six COSFIRE filters we achieve high precision P and recall R rates (P = 91.99%; R = 96.22%). This outperforms the original COSFIRE filter approach (without optimization) mostly in terms of recall. The proposed optimization procedure increases the efficiency of COSFIRE filters with little effect on the selectivity.

enjoys a high international visibility and attracts participants from all over the world.

The shape and meaning of an object can radically change with the addition of one or more contour parts. For instance, a T-junction can become a crossover. We extend the COSFIRE trainable filter approach which uses a positive prototype pattern for configuration by adding a set of negative prototype patterns. The configured filter responds to patterns that are similar to the positive prototype but not to any of the negative prototypes. The configuration of such a filter comprises selecting given channels of a bank of Gabor filters that provide excitatory or inhibitory input and determining certain blur and shift parameters. We compute the response of such a filter as the excitatory input minus a fraction of the maximum of inhibitory inputs. We use three applications to demonstrate the effectiveness of inhibition: the exclusive detection of vascular bifurcations (i.e., without crossovers) in retinal fundus images (DRIVE data set), the recognition of architectural and electrical symbols (GREC’11 data set) and the recognition of handwritten digits (MNIST data set).

The inspection of retinal fundus images allows medical doctors to diagnose various pathologies. Computer-aided diagnosis systems can be used to assist in this process. As a first step, such systems delineate the vessel tree from the background. We propose a method for the delineation of blood vessels in retinal images that is effective for vessels of different thickness. In the proposed method, we employ a set of B -COSFIRE filters selective for vessels and vessel-endings. Such a set is determined in an automatic selection process and can adapt to different applications. We compare the performance of different selection methods based upon machine learning and information theory. The results that we achieve by performing experiments on two public benchmark data sets, namely DRIVE and STARE, demonstrate the effectiveness of the proposed approach.

BACKGROUND: Keypoint detection is important for many computer vision applications. Existing methods suffer from insufficient selectivity regarding the shape properties of features and are vulnerable to contrast variations and to the presence of noise or texture. METHODS: We propose a trainable filter which we call Combination Of Shifted FIlter REsponses (COSFIRE) and use for keypoint detection and pattern recognition. It is automatically configured to be selective for a local contour pattern specified by an example. The configuration comprises selecting given channels of a bank of Gabor filters and determining certain blur and shift parameters. A COSFIRE filter response is computed as the weighted geometric mean of the blurred and shifted responses of the selected Gabor filters. It shares similar properties with some shape-selective neurons in visual cortex, which provided inspiration for this work. RESULTS: We demonstrate the effectiveness of the proposed filters in three applications: the detection of retinal vascular bifurcations (DRIVE dataset: 98.50 percent recall, 96.09 percent precision), the recognition of handwritten digits (MNIST dataset: 99.48 percent correct classification), and the detection and recognition of traffic signs in complex scenes (100 percent recall and precision). CONCLUSIONS: The proposed COSFIRE filters are conceptually simple and easy to implement. They are versatile keypoint detectors and are highly effective in practical computer vision applications.

We propose trainable filters, which we call COSFIRE (Combination Of Shifted FIlter REsponses), and use to detect vascular bifurcations in retinal images. We configure a COSFIRE filter to be selective for a bifurcation that is specified by a user in a single-step training phase. The automatic configuration comprises the selection of channels of a bank of Gabor filters and the determination of certain blur and shift parameters. A COSFIRE filter response is computed as the geometric mean of the blurred and shifted responses of the selected Gabor filters. The proposed filters share similar properties with some shape-selective neurons in visual cortex. With only five filters we achieve a recall of 98.57% at a precision of 95.37% on the 40 binary retinal images (from DRIVE), containing more than 5000 bifurcations.

We introduce a novel type of orientation–selective rank features that are sensitive to contrast modulations (second–order stimuli). Variance Ranklets are designed in close analogy with the standard Ranklets, but use the Siegel–Tukey statistics for dispersion instead of the Wilcoxon statistics. Their response shows the same orientation selectivity pattern of Haar wavelets on second–order signals that are not detectable by linear filters. To the best of our knowledge, this is the first family of rank filters designed to detect orientation in variance modulations. We validate our descriptors with an application to texture classification over a subset of the VisTex and Brodatz databases. The combination of standard (intensity) Ranklets with Variance Ranklets greatly improves on the performance of Ranklets alone. Com-parison with other published results shows that state–of–the–art recognition rates can be achieved with a simple Nearest Neighbour classifier. 1