Iceberg distribution, dispersion, and melting dynamics are pivotal in regulating the Ocean's heat and freshwater balance. However, deciphering these dynamics is a formidable challenge. In visible imagery, icebergs present significant identification challenges due to their variable appearances, which are influenced by many environmental conditions. These variations manifest as differences in color, texture, shape, and size, complicating the accurate discrimination of icebergs from open water or sea ice. Thus, developing reliable detection methods is critical for monitoring iceberg trajectories, disintegration patterns, and their consequent impact on oceanic freshwater influx. The essence of iceberg detection in visible imagery is the ability to differentiate these formations from their surrounding aquatic environment. Iceberg features display a spectrum of visual characteristics shaped by factors such as meteorological conditions, sea states, and the physical properties of the iceberg surfaces. As a result, adaptive imaging techniques are essential for efficacious detection. This study introduces an innovative Adaptive Contrast Enhancement framework meticulously crafted for iceberg detection in visible imagery. Utilizing a parameterized logarithmic model inspired by the Retinex theory, this method enhances the isolation and manipulation of image elements, thereby significantly elevating image quality. Our findings reveal that this technique markedly improves the visibility of icebergs, outshining traditional and contemporary detection methodologies. Furthermore, it affords more profound insights into the dynamic interplay of icebergs within the marine ecosystem.

Low-light image enhancement plays a crucial role for applications in security, photography, medical imaging, and scientific research. Traditional enhancement methods, including multi-spectral hardware and contrast adjustments via computer vision, often fall short due to current hardware limitations or the sparse data available in low-light conditions. This paper introduces an innovative approach that significantly improves the brightness and overall quality of low-light images, focusing on enhanced feature extraction. Our method efficiently and accurately compensates for missing data in real-time, making it highly suitable for scenarios that demand immediate processing. This is particularly beneficial for surveillance applications, where the clarity of images is essential for swift decision-making.

In this paper, we describe new methods of color image enhancement alpha-rooting by the 2D quaternion discrete Fourier transform (QDFT) in the commutative algebra, or the (2,2)-model. In this model, the concept of the convolution is unique, which is very important when transforming tasks with color images into the frequency domain. Also, there are only two types of the exponential function and therefore only two QDFTs. Both these transforms can be used to reduce the convolution to operation of multiplication. Illustrative examples on color image enhancement are given. Measures of the image enhancement and selection of the best parameters of alpha-rooting are described. A comparison with the traditional non-commutative quaternion algebra is discussed and shown that the (2,2)-model is more effective in image enhancement.

Image exploitation algorithms often require image registration to align imagery prior to processing. For example, for image fusion processing, image alignment enables consistent extraction of corresponding features. For geo-location applications, mission imagery, acquired by an onboard camera, properly-aligned with geo-located reference imagery enables extraction of geo-coordinates from pixels-of-interest. Geo-coordinates of candidate target or landmark pixels result from association with geo-located reference pixels via the image alignment. Image correlation represents a standard, classic form of image matching and image registration, which has been adapted and modified in several ways over the years. One modification consists of applying alpha-rooting to the Fourier domain image magnitudes to implement tunable Fourier domain image whitening. The whitening process emphasizes the phase content, which contains much of the image information, and mitigates effects from amplitude variations due to changing illumination conditions. Choice of alpha-rooting parameters provides the capability to adapt the whitening characteristics to the properties of the imagery, and the application in use. In this paper, we present an iterative image registration algorithm that exploits the shape properties of a region of the peak correlation coefficient, under progressive alpha-rooting parameter sharpening. Together with a coarse grid search approach, we converge to the correct image registration solution, while reducing computational load relative to a high resolution parameter space search. We provide a brief review of alpha-rooted phase correlation and describe the technical formulation. We present numerical results showing the effectiveness of the approach relative to standard correlation approaches.

Explosive detection in dual energy x-ray systems is a difficult problem owing to the fact that we don’t have enough information to estimate the effective atomic number and density of a material. Though there are several approximations available in the literature, building a solution with an acceptable true positive and false positive rate is not trivial. In this work we exploit the learning capability of a multimodal neural network for achieving a high detection rate and an acceptable false positive rate. We also show that, using a guided filter based fusion for fusing the high and low energy images leads to fused images that have a high mutual information w.r.t. the high and low images, than the existing solutions. This fused image is one of the inputs to the neural network, the other being a material dependent image that we create from the high and low energy images. The proposed solution has a high recall.

Vital signs can be inferred from facial videos for health monitoring remotely, while facial videos can be easily obtained through phone cameras, webcams, or surveillance systems. In this study, we propose a hybrid deep learning model to estimate heart rate (HR) and blood oxygen saturation level (SpO2) from facial videos. The hybrid model has a mixed network architecture consisting of convolutional neural network (CNN), convolutional long short-term memory (convLSTM), and video vision transformer (ViViT). Temporal resolution is emphasized in feature extraction since both HR and SpO2 are varying over time. A clip of video consists of a set of frame images within a time segment. CNN is performed with regard to each frame (e.g., time distributed), convLSTM and ViViT can be configured to process a sequence of frames. These high-resolution temporal features are combined to predict HR and SpO2, which are expected to capture these signal variations. Our vital video dataset is fairly large by including 891 subjects from difference races and ages. Facial detection and data normalization are performed in preprocessing. Our experiments show that the proposed hybrid model can predict HR and SpO2 accurately. In addition, those models can be extended to infer HR fluctuations, respiratory rates, and blood pressure variations from facial videos.

Support Vector Machines (SVM) have emerged as a powerful and versatile machine learning technique for solving classification and regression problems. This paper presents a thorough review of SVM, encompassing its motivation, derivation of the optimization problem, the utilization of kernels for data transformation, and a comprehensive analysis of solution methods. The review is supported by experiments conducted on a data set derived from the Traffic Sign data set. The motivation for SVM lies in its ability to address complex classification tasks by transforming the data into a higher-dimensional feature space. This is particularly beneficial for data sets derived from multiple sources. The findings presented in this paper contribute to a better understanding of SVM’s capabilities.

Breakneck advancements in aircraft technology, particularly hypersonic speeds and stealth capabilities, are challenging conventional methods of identification. These rapidly evolving capabilities make it increasingly difficult to detect and classify aircraft using traditional systems. Autonomous systems, electronic warfare capabilities, and the proliferation of unmanned aerial vehicles further develop the identification challenge. In this landscape, information becomes a vital asset, crucial for strategic decision-making, air traffic management, and safety. Accurate aircraft classification is indispensable in both military and civilian contexts. As the skies become more crowded and complex, adaptive technologies and integrating AI into identification systems become imperative to keep pace with these developments and ensure the safety and efficiency of aviation operations. In response to these challenges, we propose a novel vision transformer (ViT) designed to address the evolving landscape of aircraft identification. This ViT offers a more efficient solution through the implementation of a sparser overall structure, finely tuned for the specific application. Leveraging this model promises not only improved accuracy but also shorter training and inference times, enabling quicker and more precise aircraft classification. As we navigate the dynamic and intricate airspace of the future, this innovative ViT represents a substantial leap towards ensuring the efficacy and safety of operations.

As global populations soar and the climate warms, food supply management is an increasingly critical problem. Precision agriculture, driven by on-site data collected from various sensors, plays a pivotal role in optimizing irrigation, fertilization, and enhancing plant health and crop yield. However, the manual process of in-filed chlorophyll measurement, which is a key metric for guiding agricultural decisions, is very cumbersome and poses significant challenges. This paper explores the transformative potential of multispectral imaging data to automate plant measuring and monitoring tasks, thereby reducing labor and time costs while improving the quality of data available for making informed agricultural decisions. We present a deep-learning model for instance segmentation of plants, trained on the Growliflower dataset of RGB and multispectral image cubes of cauliflower plants. The proposed algorithm uses a Convolutional Neural Network (CNN) to leverage both the spectral information and and its spatial context to locate individual plants. We introduce a novel band-selection algorithm for determining the most significant multispectral features for use in the convolutional network: this reduces model complexity while ensuring accurate results. Our model’s ability to generalize across varying growth stages, soil conditions, and varieties of crops in the training dataset demonstrates its suitability for real-world agricultural applications. This fusion of cutting-edge sensing technology for robotic systems and state-of-the-art deep learning models holds significant promise for advancements in crop yield, resource efficiency, and sustainability practices.

As global warming causes climate change, extreme weather has become more common, posing a significant threat to life on Earth. One of the important indicators of climate change is the formation of melt ponds in the arctic region. Scarcity of large amount of annotated arctic sea ice data is a major challenge in training a deep learning model for the prediction of the dynamics of the melt ponds. In this research work, we use diffusion model, a class of generative models, to generate synthetic arctic sea ice data for further analysis of meltponds. Based on the training data, diffusion models can generate new and realistic data that are not present in the original dataset by focusing on the data distribution from a simple to a more complex distribution. First, simple distribution is transformed into a complex distribution by adding noise, such as a Gaussian distribution and through a series of invertible operations. Once trained, the model can generate new samples by starting from a simple distribution and diffusing it to the complex distribution, capturing the underlying features of the data. During inference, when generating new samples, the conditioning information is provided as input alongside the starting noise vector. This guides the diffusion process to produce samples that adhere to the specified conditions. We used high-resolution aerial photographs of Arctic region obtained during the Healy-Oden Trans Arctic Expedition (HOTRAX) in year 2005 and NASA’s Operation IceBridge DMS L1B Geolocated and Orthorectified data acquired in 2016 for the initial training of the generative model. The original image and synthetic image are assessed based on their chromatic similarity. We employed evaluation metric known as Chromatic Similarity Index (CSI) for the assessment purposes.

Marine debris is a growing threat to our oceans, impacting both wildlife and human activities. Monitoring debris, especially after natural disasters, is crucial but challenging. While Synthetic Aperture Radar (SAR) offers all-weather imaging, its effectiveness is hampered by noise and low contrast. This study proposes a new method for unsupervised SAR image enhancement using Enhanced Multilevel Enhancement (EME). EME improves image quality, allowing for more accurate debris detection compared to existing methods. This approach provides valuable insights into debris distribution across various landscapes, aiding in better understanding and managing this environmental issue.

Human physiological systems are complex nonlinear systems often exhibiting chaotic dynamics. Chaotic dynamics has been found in physiological signals of diverse types, including cardiac, respiration, and gait-based signals, and has been used to identify physiological state or diagnose abnormal physical conditions. Traditionally, contact-based sensors such as electroencephalogram, electromyogram, and electrocardiogram have been used to capture physiological signal data. However the capability to acquire physiological signals at-a-distance, offers the potential to perform remote diagnostics, health monitoring, biometrics, and activity recognition. One such non-contact sensor technology, Laser Doppler Vibrometry (LDV), captures vibration signals at offset ranges from the vibration source. Prior work has shown that LDV can capture heartbeat signals through the Doppler variations imparted by micro-vibrations of the human body due to the pulsating heart. In this paper, we investigate the detection of chaotic dynamics from real LDV cardiac signals. We interpret the cardiac signal as produced by a physiological multi-dimensional dynamical system. We first reconstruct the multidimensional phase space trajectory of the signal using the delay coordinate embedding approach, according to Takens’ theorem. We identify a de-correlation time lag using mutual information to estimate the time delay, and use the FalseNearest Neighbor approach to estimate an appropriate system dimension. We use a combination of recurrence analysis, correlation dimension, and maximum Lyapunov exponent to detect chaotic dynamics. We present numerical results demonstrating the presence of chaotic dynamics in LDV cardiac signals.

Steganography has a large range of potential applications, particularly with the advent of the internet and intellectual property concerns. A particular technique, Least Significant Bit (LSB) Steganography, is commonly used for image-in-image steganography. However, LSB steganography is a bit weak against steganalysis attacks that aim to detect the presence of embedded data. This weakness has been improved upon in Least Significant Bit Matching (LSBM) Steganography, which attempts to preserve the underlying structure of the image by maintaining a similar number of 1s and 0s as in the original image. However, standard LSBM steganography is only able to encode 1 bit of information per channel per pixel, which limits the information that could be embedded. To this end, a novel multibit approach to LSBM steganography is proposed named Multibit Least Significant Bit Matching (MLSBM) Steganography. The MLSBM approach preserves the underlying structure of the image, while allowing multiple bits to be encoded of each pixel in each channel. In addition, when the proposed MLSBM technique is used to embed high number of bits, it significantly reduces the visual perceptibility of the embedding when it is compared with the other embedding techniques for the same number of bits.

In the era of rapidly expanding image data, the demand for improved image compression algorithms has grown significantly, particularly with the integration of deep learning approaches into traditional image processing tasks. However, many of the existing solutions in this domain are burdened by computational complexity, rendering them unsuitable for real-time deployment on standard devices as they often necessitate complex systems and substantial energy consumption. This work addresses the growing paradigm of edge computing for real-time applications by introducing a novel, on-edge device solution. This innovative approach aims to strike a balance between efficiency and accuracy, adhering to the practical constraints of real-world deployment. By presenting demonstrations of the proposed solution’s performance on readily available devices, we provide tangible evidence of its applicability and viability in real-world scenarios. This advance contributes to the ongoing dialogue about the need for accessible and efficient image compression algorithms that can be deployed real-time applications on edge devices, bridging the gap between the demanding computational requirements of deep learning and the practical limitations of everyday hardware. As data continues to surge, solutions like this become ever more critical in ensuring effective image compression, aligning with on-edge computing within AI. This research paves the way for improved image processing in real-time applications while conserving computational resources and energy consumption.

Age-related macular degeneration (AMD) and Diabetic macular edema (DME) develop from irregularities in a section of the retina, causing vision impairment. Optical Coherence Tomography (OCT) imaging serves as the standard for identification, classification, and diagnosis of AMD and DME, determining locations of normal and irregular vascular patterns. However, challenges arise when OCT images are compromised by projection and motion artifacts concealing small lesions. This paper aims to develop an automated system for quantifying and categorizing AMD and DME (diabetic macula edema). The proposed approach, Multi-Kernel Wiener Local Binary Patterns (MKW-LBP) uses kernels of various sizes for feature extractions. Our proposed method is twofold: (1) Wiener patterns extract retinal features, robust against motion artifacts, thus preserving lesion visibility, and (2) multi-kernel vectorization exploits textural feature. Computer simulations demonstrate that the proposed technique achieves an overall accuracy of 98% through ten-fold cross-validation on the Duke University dataset. Furthermore, our system exhibits strong resistance against added Gaussian Noise, ensuring reliable performance under severe noise.

Wireless capsule endoscopy (WCE) offers a minimally invasive approach to inspecting the gastrointestinal (GI) tract, crucial for diagnosing conditions such as malnutrition, dehydration, and potential cancers. However, WCE image diagnostics can be compromised by inadequate illumination and adversarial contrast reduction attacks. Adversarial contrast reduction attacks are intentional efforts to degrade image contrast and mislead automated diagnosis systems. Such challenges can result in misclassifications, negatively impacting patient safety. This study examines the effects of contrast degradation on Deep Learning (DL) models designed for WCE image analysis. The study emphasizes the adverse impact of substantial contrast reductions from adversarial attacks on classification accuracy. We propose a novel texture descriptor to mitigate this vulnerability: the Color Quaternion Modulus and Phase Patterns (CQ-MPP). This descriptor effectively extracts textural features within WCE images, enabling the identification of potentially cancerous regions, even under significantly reduced contrast. The effectiveness of CQ-MPP is evaluated using the Wireless Capsule Endoscopy Curated Colon Disease Dataset. Results show that CQ-MPP maintains good accuracy in detecting cancerous lesions and demonstrates remarkable resilience to contrast adversarial degradation. This method ensures reliable performance amidst severe contrast reduction, offering significant potential to improve safety of GI disease diagnosis via WCE.

The quantum Fourier transform (QFT) is the important operation in quantum computing. It is used in many algorithms, including the Shor's quantum algorithm for finding the prime factors of integers. Color image encryption, processing and representation of quantum images are areas where the QFT is also used. In this paper, we discuss the quantum superpositions of the images and methods of calculation of the analogues of the quantum 2-D discrete Fourier transform of images. The quantum algorithms and circuits will be described, and examples of calculation of the 2-D QFT of images of 8×8 and 16×16 pixels will be given in detail.

Images and videos captured in poor illumination conditions are degraded by low brightness, reduced contrast, color distortion, and noise, rendering them barely discernable for human perception and ultimately negatively impacting computer vision system performance. These challenges are exasperated when processing video surveillance camera footage, using this unprocessed video data as-is for real-time computer vision tasks across varying environmental conditions within Intelligent Transportation Systems (ITS), such as vehicle detection, tracking, and timely incident detection. The inadequate performance of these algorithms in real-world deployments incurs significant operational costs. Low-light image enhancement (LLIE) aims to improve the quality of images captured in these unideal conditions. Groundbreaking advancements in LLIE have been recorded employing deep-learning techniques to address these challenges, however, the plethora of models and approaches is varied and disparate. This paper presents an exhaustive survey to explore a methodical taxonomy of state-of-the-art deep learning-based LLIE algorithms and their impact when used in tandem with other computer vision algorithms, particularly detection algorithms. To thoroughly evaluate these LLIE models, a subset of the BDD100K dataset, a diverse real-world driving dataset is used for suitable image quality assessment and evaluation metrics. This study aims to provide a detailed understanding of the dynamics between low-light image enhancement and ITS performance, offering insights into both the technological advancements in LLIE and their practical implications in real-world conditions.

Recent progress in deraining and dehazing methods has dramatically enhanced image quality in bad weather. However, these methods are vulnerable to adversarial attacks, severely compromising their effectiveness. Traditional defenses like adversarial training and model distillation necessitate significant retraining, hindering their real-world application due to high computational costs. To address these limitations, we propose the Quaternion-Hadamard Transformer Network (QHTN), a novel defense strategy against white-box adversarial attacks, including the Fast Gradient Sign Method (FGSM) and Projected Gradient Descent (PGD). The QHTN leverages a transformer architecture with three key modules: preprocessing, local-global feature extraction, and reconstruction. The local-global feature extraction module utilizes innovative Hadamard and quaternion convolution blocks to analyze spatial and inter-channel relationships. This unique approach enables the QHTN to incorporate a denoising mechanism during preprocessing, effectively mitigating adversarial noise before it influences the model's input. Extensive evaluations demonstrate the QHTN's efficacy in safeguarding haze and rain removal models from adversarial attacks. These results validate the QHTN's efficiency and potential for broader adoption in image-processing defense mechanisms.

To advance road safety through technological innovation, we present the Comprehensive Urban Navigation and Yielding (CUNY) Video Dataset (CVD), a pioneering collection aimed at enriching the analysis of roadway incidents using stationary camera footage. Derived from 1,013 YouTube videos, CVD is intricately annotated to discern between collision and non-collision scenarios, opening avenues for profound insights into various roadway incidents. CVD has been meticulously curated to overcome prevalent limitations in existing collision databases, boasting a comprehensive representation of environmental conditions, camera qualities, geographical diversity, and temporal variations. It is particularly well-suited for integration with existing road monitoring infrastructures, enabling optimization of emergency response, enhancement of traffic management, and overall improvement in road safety. By openly disseminating this dataset, we seek to address the scarcity of accessible, diverse, and authentic video data for collision analysis, contributing to advancements in the field of intelligent transportation systems and fostering safer road environments.

In the digital age, web apps are vital for global communication and commerce, necessitating robust security measures. This study delves into Automated Web Application Penetration Testing (AWAPT) tools, assessing their effectiveness against the complexities of modern web technologies. It highlights the critical need for a nuanced analysis of these tools, considering their adaptability, accuracy, coverage, ease of use, and flexibility. The aim is to offer practical advice for selecting appropriate tools for diverse web applications, addressing the increasing cyber threats and reliance on web apps. The paper identifies a gap in aligning current tools with advanced web technologies and the lack of comprehensive evaluations, posing risks to web app security. It calls for future research on evolving technologies, tool effectiveness, and advanced techniques like AI to enhance tool robustness against new threats. The study's comparative analysis seeks to benchmark tool performance, identifying strengths and weaknesses to improve their effectiveness in securing web applications against modern challenges.

This paper explores the deployment of digital voting systems in democracies, focusing on blockchain technology as a solution to security, integrity, and public trust challenges. It delves into blockchain's fundamentals, its integration into e-voting systems, technical attributes, societal impacts, regulatory compliance, and a proof of concept. By analyzing blockchain's ability to protect against electoral fraud and manipulation through security techniques like cryptography, the study highlights its potential to ensure secure, transparent elections. The research includes an in-depth examination of societal acceptance, technical solutions, and legal frameworks based on European experiences, proposing a system architecture that ensures voter anonymity and vote integrity. The development of a blockchain-based prototype illustrates the feasibility of such systems. This comprehensive analysis offers insights into technical and societal considerations for digital voting, suggesting blockchain technology as a revolutionary tool for enhancing democratic participation and governance.

The 300th birthday of the world-famous German philosopher Immanuel Kant (1724-1804) in 2024 offers an extraordinary opportunity to explore integrating advanced technologies into cultural mega-events. This paper examines the opportunities and challenges of using various technologies, such as Extended Reality (XR) and Visual Effects (VFX), to bring this essential European intellectual to a broader audience. These technologies, including technological advances based on real-time rendering in conjunction with LED volumes and high-quality 3D assets, offer immersive and interactive event experiences. However, the use of these technologies presents specific challenges, such as processing limitations and the need to create specialized content, including the development of 3D models relevant to the life and philosophy of Immanuel Kant. This article provides an overview of the current state of such opportunities and challenges and examines their application in the context of this cultural mega-event. The aim is to create a compelling event experience, e.g., as part of a museum visit, while simultaneously using technology in a meaningful educational way. It is essential not to overload historical and intellectual personalities such as Immanuel Kant with technological possibilities in the sense of an ethical responsibility. At the same time, the creation and communication of knowledge require explanation in a museum context or another explanatory framework.

The proliferation of digital medical imaging has ushered in unprecedented advancements in healthcare diagnostics and treatment planning. However, this digital era has also raised significant concerns regarding the security and privacy of sensitive medical image data, particularly in the context of the Digital Imaging and Communications in Medicine (DICOM) standard. In response to these concerns, this paper presents a novel Chaos-Driven Encryption (CDE) system designed to fortify the confidentiality and integrity of DICOM medical images. The CDE system harnesses chaotic systems' inherent unpredictability and complexity to create a robust encryption framework. Chaos-based encryption offers a formidable defence against conventional cryptographic attacks due to its nonlinearity and sensitivity to initial conditions. We propose a specific chaotic map and key management scheme tailored for DICOM images, ensuring that the encryption process remains secure and efficient. This paper comprehensively analyzes the CDE system, including its encryption process, key generation, and decryption procedure. We assess the security of CDE through extensive cryptographic analysis, demonstrating its resistance to known attacks and vulnerabilities. Moreover, we evaluate the computational performance of CDE in terms of encryption/decryption speed and resource utilization. Our experimental results highlight the feasibility and effectiveness of Chaos-Driven Encryption in safeguarding DICOM medical images against unauthorized access and tampering. By presenting this innovative encryption system, we contribute to the ongoing dialogue on healthcare data security and privacy, offering a promising solution for the protection of sensitive medical image information in the digital age.

In underwater exploration, Autonomous Underwater Vehicles (AUVs) face challenges due to the adverse effects of the aquatic environment on optical sensors, resulting in sub-optimal data acquisition. To overcome this, we propose a novel solution utilizing a Generative Adversarial Network (GAN) model. Rooted in the U-Net architecture, our model processes low-quality AUV camera feed, generating enhanced representations of the underwater scene. The discriminator focuses on evaluating current image patches, capturing high-frequency properties with fewer parameters, achieving a 15% improvement in model accuracy. This approach facilitates realtime preprocessing in visually-guided underwater robot autonomy pipelines, overcoming challenges associated with underwater visibility

Infrared and thermal images have been used widely in different security applications. One of the drawbacks of such images is low contrast and noisy images, which should be enhanced. We present a new image enhancement algorithm based on block-rooting processing with artificial multi-scale-exposure image fusion. The proposed block-based multi-scale enhancement method is based on a 3-D block-rooting transform domain technique comprised: finding similar blocks in the image by block-matching; block-grouping for different block sizes; applying 3-D block-matching image enhancement; decomposition of the weight map and multi-scale enhanced images into the Gaussian and Laplacian pyramids; fusion by multiplying multi-scale images and weights. A new stage is proposed to obtain a local-global estimate of high-contrast images, also used in the general artificial fusion model. Some presented experimental results illustrate the performance of the proposed method on the thermal image dataset compared with the traditional methods.