Set Of Pixels Research Articles

Semi-supervised learning for concrete defect segmentation from images

Deep learning-based concrete defect detection has emerged as a promising technology to overcome the limitations of manual visual inspection, which is often subjective, laborious, and time-consuming. However, most existing methods rely heavily on manual labels and adopt a fully supervised learning approach, which hinders their practical application in engineering. To tackle these concerns, we proposed a semi-supervised learning model, termed Semi-SegDefect. Our proposal utilizes a mean teacher architecture to assign pseudo-labels to the pixels of unlabeled images and performs pixel-level contrastive learning on a sparse set of hard negative pixels to achieve segmentation boundaries with the highest possible accuracy. Test results demonstrate that Semi-SegDefect outperforms fully supervised models significantly, especially when trained with limited labeled data. This method shows great promise for enhancing the accuracy and scalability of concrete defect segmentation and can contribute to the advancement of the construction industry.

Robust Image Steganography Against General Downsampling Operations With Lossless Secret Recovery

Resisting the operations in lossy channels is a challenge for image steganography. In this paper, we propose a novel robust steganographic method to resist the image downsampling operation. Unlike the existing schemes, our method guarantees lossless secret recovery from the stego-image after general image downsampling operations, which considers the undetectability of the stego-images on both sides (sender and receiver) of the lossy channel. We first downsample the cover image to get its downsampled version on the receiver side, and select a set of embeddable pixels (i.e., the pixels that can be modified for data embedding) from the downsampled image. Then, we generate a stego-image on the receiver side (termed as the receiver stego-image) such that the distortion caused by the data embedding is minimized. Based on the receiver stego-image, we modify the cover image to produce the stego-image on the sender side (termed as the sender stego-image). The modification takes the embedding cost into account and makes sure that the downsampled version of the sender stego-image produces the same embeddable pixels as the receiver stego-image. Experimental results show that our method performs significantly better than the existing schemes in terms of robustness and undetectability for resisting general image downsampling operations.

Interactive experience of sculpture design based on virtual reality technology

Abstract The application of computer technology provides new ideas for the design of sculpture works. In this paper, with the support of virtual reality technology, the pixel set of each color block is extracted from the actual image to determine the location of the center point of the color block. Based on the extracted information, a visual color marking method is used to realize the 3D registration of the sculpture scene. In the virtual scene, through the virtual modeling of the sculpture work, a spline surface model was generated by applying sparse image sequences and optimally adjusted to transform the surface piece control point problem into a quadratic function of the network vertices. The mathematical logic properties of the virtual reality sculpture work wind direction improved by 22.5 points, the mathematical logic properties of the light design improved by 25.8 points, and the mathematical logic properties of the shadow design improved by 23.1 points. The virtual reality technology provides an interactive experience for the design of sculpture works.

An adversarial attack approach for eXplainable AI evaluation on deepfake detection models

With the rising concern on model interpretability, the application of eXplainable AI (XAI) tools on deepfake detection models has been a topic of interest recently. In image classification tasks, XAI tools highlight pixels influencing the decision given by a model. This helps in troubleshooting the model and determining areas that may require further tuning of parameters. With a wide range of tools available in the market, choosing the right tool for a model becomes necessary as each one may highlight different sets of pixels for a given image. There is a need to evaluate different tools and decide the best performing ones among them. Generic XAI evaluation methods like insertion or removal of salient pixels/segments are applicable for general image classification tasks but may produce less meaningful results when applied on deepfake detection models due to their functionality. In this paper, we perform experiments to show that generic removal/insertion XAI evaluation methods are not suitable for deepfake detection models. We also propose and implement an XAI evaluation approach specifically suited for deepfake detection models.

MAPMaN: Multi‐Stage U‐Shaped Adaptive Pattern Matching Network for Semantic Segmentation of Remote Sensing Images

AbstractRemote sensing images (RSIs) often possess obvious background noises, exhibit a multi‐scale phenomenon, and are characterized by complex scenes with ground objects in diversely spatial distribution pattern, bringing challenges to the corresponding semantic segmentation. CNN‐based methods can hardly address the diverse spatial distributions of ground objects, especially their compositional relationships, while Vision Transformers (ViTs) introduce background noises and have a quadratic time complexity due to dense global matrix multiplications. In this paper, we introduce Adaptive Pattern Matching (APM), a lightweight method for long‐range adaptive weight aggregation. Our APM obtains a set of pixels belonging to the same spatial distribution pattern of each pixel, and calculates the adaptive weights according to their compositional relationships. In addition, we design a tiny U‐shaped network using the APM as a module to address the large variance of scales of ground objects in RSIs. This network is embedded after each stage in a backbone network to establish a Multi‐stage U‐shaped Adaptive Pattern Matching Network (MAPMaN), for nested multi‐scale modeling of ground objects towards semantic segmentation of RSIs. Experiments on three datasets demonstrate that our MAPMaN can outperform the state‐of‐the‐art methods in common metrics. The code can be available at https://github.com/INiid/MAPMaN.

Generating Virtual Training Labels for Crop Classification from Fused Sentinel-1 and Sentinel-2 Time Series

Convolutional neural networks (CNNs) have shown results superior to most traditional image understanding approaches in many fields, incl. crop classification from satellite time series images. However, CNNs require a large number of training samples to properly train the network. The process of collecting and labeling such samples using traditional methods can be both, time-consuming and costly. To address this issue and improve classification accuracy, generating virtual training labels (VTL) from existing ones is a promising solution. To this end, this study proposes a novel method for generating VTL based on sub-dividing the training samples of each crop using self-organizing maps (SOM), and then assigning labels to a set of unlabeled pixels based on the distance to these sub-classes. We apply the new method to crop classification from Sentinel images. A three-dimensional (3D) CNN is utilized for extracting features from the fusion of optical and radar time series. The results of the evaluation show that the proposed method is effective in generating VTL, as demonstrated by the achieved overall accuracy (OA) of 95.3% and kappa coefficient (KC) of 94.5%, compared to 91.3% and 89.9% for a solution without VTL. The results suggest that the proposed method has the potential to enhance the classification accuracy of crops using VTL.

A Machine Learning Approach for Detection and Suppression of Shadow or Wet Road Surfaces

In advanced driver assistance system detection of road surfaces is an important task. Few algorithms have been proposed in past to detect the road surfaces based on intensities. However, problem arises in detection process is due to the presence of shadows or wet road surfaces. Here we have proposed a novel algorithm for detection of shadows with the help of machine learning approaches. Initially shadow is being detected with the help of a threshold-based approach followed by windowing-based method. The detected shadow region gets confirmed with the help of a set of features and classifier. The detected shadow or wet pixels are in painted to obtain set of pixels without shadow for road classification problems. The simplicity and accuracy of the algorithm makes it robust and can be used as a part of road surface detection algorithm.

Dynamic contrastive learning guided by class confidence and confusion degree for medical image segmentation

This work proposes an intra-Class-confidence and inter-Class-confusion guided Dynamic Contrastive (CCDC) learning framework for medical image segmentation. A core contribution is to dynamically select the most expressive pixels to build positive and negative pairs for contrastive learning at different training phases. For the positive pairs, dynamically adaptive sampling strategies are introduced for sampling different sets of pixels based on their hardness (namely the easiest, easy, and hard pixels). For the negative pairs, to efficiently learn from the classes with high confusion degree w.r.t a query class (i.e., a class containing the query pixels), a new hard class mining strategy is presented. Furthermore, pixel-level representations are extended to the neighbourhood region to leverage the spatial consistency of adjacent pixels. Extensive experiments on the three public datasets demonstrate that the proposed method significantly surpasses the state-of-the-art.

A Boosting Fuzzy Clustering Technique for Image Segmentation based on an Initial Low-Class

Abstract: Image segmentation is a basic problem in medical image analysis and useful for disease diagnosis. However, the complexity of medical images makes image segmentation difficult. In recent decades, fuzzy clustering algorithms have been preferred due to their simplicity and efficiency. However, they are sensitive to noise. To solve this problem, many algorithms using non-local information have been proposed, which perform well but are inefficient. This paper proposes a boosting fuzzy clustering algorithm utilizing non- local self-similarity and a low-class prior for image segmentation. Firstly, cluster centers are initialized based on peak detection. Then, a pixel correlation model between corresponding pixels is constructed, and similar pixel sets are retrieved. To improve efficiency and robustness, the proposed algorithm uses a novel objective function combining non-local information and an initial low-class. The algorithm's ability to achieve satisfactory results while simultaneously significantly increasing efficiency is demonstrated by experiments on medical and synthetic images.

Algebraic Multi-Layer Network: Key Concepts.

The paper refers to interdisciplinary research in the areas of hierarchical cluster analysis of big data and ordering of primary data to detect objects in a color or in a grayscale image. To perform this on a limited domain of multidimensional data, an NP-hard problem of calculation of close to optimal piecewise constant data approximations with the smallest possible standard deviations or total squared errors (approximation errors) is solved. The solution is achieved by revisiting, modernizing, and combining classical Ward's clustering, split/merge, and K-means methods. The concepts of objects, images, and their elements (superpixels) are formalized as structures that are distinguishable from each other. The results of structuring and ordering the image data are presented to the user in two ways, as tabulated approximations of the image showing the available object hierarchies. For not only theoretical reasoning, but also for practical implementation, reversible calculations with pixel sets are performed easily, as with individual pixels in terms of Sleator-Tarjan Dynamic trees and cyclic graphs forming an Algebraic Multi-Layer Network (AMN). The detailing of the latter significantly distinguishes this paper from our prior works. The establishment of the invariance of detected objects with respect to changing the context of the image and its transformation into grayscale is also new.

THE METHOD OF SEGMENTATION OF THE IMAGES IN THE REMOTE VIDEO MONITORING SYSTEM BASED ON THE K-MEANS ALGORITHM

The configuration of the video monitoring system, which is widely used in the protection systems of objects and positions of units, is given. Examples of the collision line video monitoring system are given, namely: the video monitoring system of the special monitoring mission of the Organization for Security and Cooperation in Europe and the battalion video surveillance complex. An analysis of known methods of image segmentation in video monitoring systems has been carried out. The analysis showed that these methods do not meet the requirements for the quality of image segmentation in the video monitoring system. The image segmentation method in the remote video monitoring system has been improved, in which, unlike the known ones: the source image is divided into RGB brightness channels; determination of the Euclidean distance between pixels; division of the entire set of image pixels into subsets; recalculation of “centers” of each subset; reassignment of new “centers” of each subset; minimization of total intra-cluster variance. Experimental studies of image segmentation from the remote video monitoring system were conducted. The image from the video monitoring system on the contact line during hostilities during the russian-ukrainian war is used as the source image. The segmentation results of this source image are given. It was established that there is a visual possibility of identifying objects of interest on segmented images. At the same time, it is possible to visually determine the elements of the protective cape only at k=4, while at k=2 and k=3 this possibility is complicated. The directions of further research are the determination of the optimal value of the value k when using the k-means algorithm and the evaluation of the quality of image segmentation in the remote video monitoring system.

Determination of the number of clusters on images from space optic-electronic observation systems using the k-means algorithm

The object of research is the process of clustering images from space optical-electronic surveillance systems. The main hypothesis of the study assumed that experimental studies would make it possible to determine the number of clusters on images from space optical-electronic surveillance systems when using the k-means algorithm. The method of clustering images from space optical-electronic surveillance systems using the k-means algorithm, unlike the known ones, implies: – splitting the source image into Red-Green-Blue brightness channels; – determination of the Euclidean distance between pixels; – distribution of the entire set of image pixels into clusters; – recalculation of "centers" of each subset; – reassignment of new "centers" of each cluster; – minimization of the total intracluster variance. Experimental studies were conducted on the clustering of the original image using the k-means method at different values of k. It was established that with an increase in the value of k, the visual quality of clustering improves, and it is possible to visually determine a larger number of clusters in the images. To determine the number of clusters, the sum of clustering errors of type 1 and 2 at different values of k was evaluated. It was established that when the value of k increases, the sum of errors of the 1st and 2nd kind initially decreases exponentially. A further increase in the value of k does not lead to a significant decrease in errors of the 1st and 2nd kind. It was established that for a typical image from the space optical-electronic observation system, the value of k in the clustering method based on the k-means algorithm should be equal to 4. At the same time, the sum of errors of the 1st and 2nd kind is 31.3 %. Further research is directed to the development of clustering methods that reduce the sum of errors of the 1st and 2nd kind

Modelling Spectral Unmixing of Geological Mixtures: An Experimental Study Using Rock Samples

Spectral unmixing of geological mixtures, such as rocks, is a challenging inversion problem because of nonlinear interactions of light with the intimately mixed minerals at a microscopic scale. The fine-scale mixing of minerals in rocks limits the sensor’s ability to identify pure mineral endmembers and spectrally resolve these constituents within a given spatial resolution. In this study, we attempt to model the spectral unmixing of two rocks, namely, serpentinite and granite, by acquiring their hyperspectral images in a controlled environment, having uniform illumination, using a laboratory-based imaging spectroradiometer. The endmember spectra of each rock were identified by comparing a limited set of pure hyperspectral image pixels with the constituent minerals of the rocks based on their diagnostic spectral features. A series of spectral unmixing paradigms for explaining geological mixtures, including those ranging from simple physics-based light interaction models (linear, bilinear, and polynomial models) to classification-based models (support vector machines (SVMs) and half Siamese network (HSN)), were tested to estimate the fractional abundances of the endmembers at each pixel position of the image. The analysis of the results of the spectral unmixing algorithms using the ground truth abundance maps and actual mineralogical composition of the rock samples (estimated using X-ray diffraction (XRD) analysis) indicate a better performance of the pure pixel-guided HSN model in comparison to the linear, bilinear, polynomial, and SVM-based unmixing approaches. The HSN-based approach yielded reduced errors of abundance estimation, image reconstruction, and mineralogical composition for serpentinite and granite. With its ability to train using limited pure pixels, the half-Siamese network model has a scope for spectrally unmixing rock samples of varying mineralogical composition and grain sizes. Hence, HSN-based approaches effectively address the modelling of nonlinear mixing in geological mixtures.

Analysis of Land Use and Land Cover Changes in Owerri Municipal and Its Environs, Imo State, Nigeria (2005 – 2015)

Land use/land cover changes in Owerri Municipal and its environs which included Orogwe, Ubomiri, Emii, Uratta, Ihiagwa and Egbu were studied as being driven by rapid population growth for urbanisation. The study applied the digital technology of Remote Sensing and Geographic Information System (ARC GIS 9.3 Version software). These tools were implored to generate land use/land cover maps for the period; 2005, 2010 and 2015 and to determine the area in km2 of each of the six classes of land use/land cover types, the percentage change of the total area covered, accuracy of the overall classification including the Kappa coefficient, while a classification scheme was used to develop the classified land use/land cover maps. Also satellite imagery for the period (2005-2015) was imputed into the ENVI 4.5 software environment, composited, digitized and exported to the Arc. GIS where they were clipped with the study area. The extracted image of the study area was then exported back to the ENVI 4.5 environment for Arc.GIS environment in TIFF format. This was followed by a colour separation in the imagery repeated for all the raw satellite imagery. Image interpretation was done on ENVI 4.5 software based on a set of pixels of the Region Of Interest (ROI). Results are that Kappa coefficient values were high enough for the period of study with 0.9099, 0.9557 and 0.9685 for 2005, 2010 and 2015 respectively indicating a strong agreement between the classified maps and ground referenced information. Integrating GIS and satellite remote sensing with high spectral, spatial and temporal resolution at the local scale to develop urban environmental monitoring, effective land use planning and management of the current growth pattern were among others recommended.

REAL-TIME PLOTTING OF BIOMEDICAL SIGNALS IN APPLICATIONS UTILIZING WINDOWS PRESENTATION FOUNDATION TECHNOLOGY

Abstract: Utilization of personal computers in medicine requires strong emphasis on software performance. Contemporary programming languages give engineers great flexibility in developing applications. Since these technologies are more and more complex, there is a need to analyze methods of achieving desired results taking into account time requirements. This paper focuses mainly on development of plotting module using Windows Presentation Foundation library, which is one of the newest tool for software developers. Aspects of plotter creation for technologically limited screen are covered and conclusions are summarized giving recommendations regarding the topic. General methods of drawing signal on a raster surface are also discussed. Since medical signals in monitoring range consist of frequencies up to about 100 Hz, one has to take into account aspects of signal scaling in order to not to make its important parts invisible. The main problem is that personal computer screen, e.g. a monitor, is purely digital in nature; actually it is the set of pixels (consisting of three components). Because resolution is limited, the signal must be limited in frequency to the range which can be shown at a given plotting speed. This paper describes these aspects of signal plotting as well as the method for virtually increasing pixel resolution by using particular components of a pixel. This paper is a summary giving a general idea of the capabilities of the above mentioned WPF technology and the basis for creating such programs using methods such as: WriteableBitmap method, DrawingVisual method, Canvas Method type 1, Canvas Method type 2.

Deep Learning Robustness Verification for Few-Pixel Attacks

While successful, neural networks have been shown to be vulnerable to adversarial example attacks. In L 0 adversarial attacks, also known as few-pixel attacks, the attacker picks t pixels from the image and arbitrarily perturbs them. To understand the robustness level of a network to these attacks, it is required to check the robustness of the network to perturbations of every set of t pixels. Since the number of sets is exponentially large, existing robustness verifiers, which can reason about a single set of pixels at a time, are impractical for L 0 robustness verification. We introduce Calzone, an L 0 robustness verifier for neural networks. To the best of our knowledge, Calzone is the first to provide a sound and complete analysis for L 0 adversarial attacks. Calzone builds on the following observation: if a classifier is robust to any perturbation of a set of k pixels, for k > t , then it is robust to any perturbation of its subsets of size t . Thus, to reduce the verification time, Calzone predicts the largest k that can be proven robust, via dynamic programming and sampling. It then relies on covering designs to compute a covering of the image with sets of size k . For each set in the covering, Calzone submits its corresponding box neighborhood to an existing L ∞ robustness verifier. If a set’s neighborhood is not robust, Calzone repeats this process and covers this set with sets of size k ′< k . We evaluate Calzone on several datasets and networks, for t ≤ 5. Typically, Calzone verifies L 0 robustness within few minutes. On our most challenging instances (e.g., t =5), Calzone completes within few hours. We compare to a MILP baseline and show that it does not scale already for t =3.

Chains Defects Detection in a Printed Circuit Board Image by the Plane Partition and Flood-filling of Traces

An approach to dividing the printed circuit board into parts to increase visibility of defects in a PCB image is considered. The approach is based on a piecewise linear approximation of a cumulative histogram. The last one is calculated for numbers of informative pixels placed in rows and columns of an image matrix. Informative pixels are those indicating contacts and C traces. The histogram demonstrates a distribution of informative pixels along the OX and OY axes. The beginning and ending points of linear lines are taken as coordinates of the divided parts of the PCB board. The flood-fill algorithm is used to color and separate the PCB chains components. The start pixels are taken from a set of informative pixels. To measure defect values, a mean intensity and gradient functions for the template and manufactured images are used. Distributed cumulative histograms are applied to PCB components to detect places and intensity of defects.

Experimental Implementation of Metasurfaces for Secure Multi‐Channel Image Encryption in the Infrared

AbstractThe ability to tailor light–matter interactions using artificially engineered materials has opened up new avenues for secure data storage and communication. This work presents an experimental investigation of metasurfaces for secure, multi‐channel image encryption in the infrared (IR). The proposed metasurfaces consist of an array of pixels, each designed to produce a wavelength‐ and polarization‐dependent IR absorptivity. A basis set of pixels is designed for encrypting images of arbitrary resolution on a given number of wavelength and polarization channels. These pixels are fabricated, and their spectral response is experimentally measured using Fourier transform infrared spectroscopy. The measured data is used to emulate the encryption and decryption of binary and 8‐bit grayscale images. Finally, the security of the encryption scheme proposed in this work is evaluated by performing statistical analyses on the image data stored on different channels. The results presented in this study suggest intriguing possibilities for the development of encrypted tagging technologies in the infrared and thus have implications for secure object identification and anti‐counterfeiting.

Evaluation of a correlation phase recovery method to be used in temporal speckle pattern interferometry with noise contaminated specklegrams

When a phase distribution is evaluated in temporal speckle pattern interferometry, the accuracy of the retrieved phase distribution depends on how the different phase recovery methods deal with the influence of low-frequency noise and intensity bias. In this paper, we show that the proposed correlation method can effectively recover the phase distribution in time series of speckle interferograms that are affected by sets of adjacent non-modulated pixels, combined with very small phase changes between successive frames. The performance of the proposed method is analyzed when the retrieved optical phase distributions are coded in noisy temporal speckle pattern interferometry signals presenting large regions of adjacent low-modulated pixels. It is also compared with the data measured with the well-known one-dimensional Fourier transform technique, and the three-dimensional directional wavelet transform method. The performance of the proposed phase retrieval technique is compared with these last two methods using computer-simulated speckle interferograms, and the advantages and limitations of the proposed method are discussed. Finally, the performance of the proposed correlation technique to process real data is illustrated.

About Search Synthesis of the Predictor for Image Compression

In the field of lossless halftone image compression, high compression results are provided by a predictive compression scheme. One predictor is the GAP (Gradient Adjusted Prediction) predictor. This article explores the possibility of predictor search synthesis, like GAP predictor, considers the possibility of search synthesis of optimal predictor model parameters for each possible GAP predictor context, while the test images were divided into contexts with sets of neighborhood pixels. When implementing the search synthesis of a predictor, like the GAP predictor, a classical genetic algorithm was used, however, in order to improve the operation of the classical genetic algorithm, a study was carried out on pseudo-random number generators with a uniform distribution, built using the multiplicative sensor method. The synthesis of pseudo-random number generators with a uniform distribution was automated, the statistical hypothesis on the type of distribution provided by the pseudo-random number generator is automatically tested, for this the value of the Chi-square distribution is calculated, a simplified calculation of the Gamma function and numerical integration according to the Simpson formula is performed. As a result, pseudo-random number generators were not built into the programming tools, but selected on their own, the periods of selected pseudo-random number generators are in the range from 402 544 072 to 2 544 344 962. The discrete nature of defining GAP contexts in the original GAP leads to a noticeable overestimation of the entropy of the prediction error. Thus, for context 32, the difference between entropy achieved as a result of optimization of model parameters and entropy achieved by the original GAP predictor is 0.1148 bpp, for context 33 - 0.02568 bpp, for context 8 - 0.07011 bpp, for context 9 - 0.00925 bpp. The GAP predictor does not have many possibilities to improve prediction by determining optimal predictor performance parameters for each possible context. However, due to the low computational complexity of the GAP predictor, it is important to investigate opportunities to improve the effectiveness of the GAP predictor.