Multimedia Tools and Applications最新文献

Retinal vessel segmentation has various applications in the biomedical field. This includes early disease detection, biometric authentication using retinal scans, classification and others. Many of these applications rely critically on an accurate and efficient segmentation technique. In the existing literature, a lot of work has been done to improve the accuracy of the segmentation task, but it relies heavily on the amount of data available for training as well as the quality of the images captured. Another gap is observed in terms of the resources used in these heavily trained algorithms. This paper aims to address these gaps by using a resource-efficient unsupervised technique and also increasing the accuracy of retinal vessel segmentation using the Fourier decomposition method (FDM) along with the Gabor transform for image signals. The proposed method has an accuracy of 97.39%, 97.62%, 95.34%, and 96.57% on DRIVE, STARE, CHASE_DB1, and HRF datasets, respectively. The sensitivities were found to be 88.36%, 88.51%, 90.37%, and 79.07%, respectively. A separate section makes a detailed comparison of the proposed method with several well-known methods and an analysis of the efficiency of the proposed method. The proposed method proves to be efficient in terms of time and resource requirements.

This paper introduces the innovative concept of the Cubixel—a three-dimensional representation of the traditional pixel—alongside the derived metric, Volume of the Void (VoV), which measures spatial disparities within images. By converting pixels into Cubixels, we can analyze the image’s 3D properties, thereby enriching image processing and computer vision tasks. Utilizing Cubixels, we’ve developed algorithms for advanced image segmentation, edge detection, texture analysis, and feature extraction, yielding a deeper comprehension of image content. Our empirical experimental results on benchmark images and datasets showcase the applicability of these concepts. Further, we discuss future applications of Cubixels and VoV in various domains, particularly in medical imaging, where they have the potential to significantly enhance diagnostic processes. By interpreting images as complex ‘urban landscapes’, we envision a new frontier for deep learning models that simulate and learn from diverse environmental conditions. The integration of Cubixels into deep learning architectures promises to revolutionize the field, providing a pathway towards more intelligent, context-aware artificial intelligence systems. With this groundbreaking work, we aim to inspire future research that will unlock the full potential of image data, transforming both theoretical understanding and practical applications. Our code is available at https://github.com/sanadv/Cubixel.

The desire for new algebraic structures is always an interesting area of research for the design and development of new information confidentiality mechanisms. Confidentiality mechanism keeps information secure from unauthentic users and transforms information into a ciphered format that cannot be easily deciphered. The key components of data security are encryption/decryption algorithms. Encryption/Decryption Algorithms depend on mathematical structures used for substitution box (S-box) and key scheduling. In the present research, the Galois ring-based S-box is used in the Blowfish confidentiality technique and practically implement for electro-optical satellite image ciphering. The results of data ciphering are compared with Blowfish cipher with the standard S-box. Ciphered images have been evaluated by standard tests such as histogram equalization, randomness, correlation, and differential assaults analysis. These tests depict that the ciphered data by Blowfish based on Galois ring-based S-box is more secure and the security is achieved in the fourth round for electro-optical satellite images. Test results for satellite images illustrate that by less execution time found better satellite image ciphering in the fourth ciphering round by using the proposed scheme. With this addition, subsequent steps of the existing algorithm were modified which add more robustness to our digital information. Security analysis of this tested mechanism added more robustness to the confidentiality of digital images. The results of the anticipated ciphering scheme clarify the better performance as compared to the standard Blowfish algorithm for satellite images.

The COVID-19 pandemic was triggered by the SARS-CoV-2 virus which caused multiple ill-health conditions in infected individuals. There were many cases that culminated in death. Chest X-ray images became a proven method for spotting thoracic ailments. The resultant availability of huge public datasets of chest X-ray images has great potential in deep learning for lung ailment detection. This paper presents a classification that aims at acquiring the optimal hyperparameters using the metaheuristic algorithm for various pre-trained CNN training processes. The experimental results show that HSAGWO (Hybrid Simulated Annealing Grey Wolf Optimization) outperforms the other contemporary models for optimizing training hyperparameters in the ResNet50 network. The accuracy, precision, sensitivity (recall), specificity, and F1-score values obtained are 98.78%, 98.10%, 99.31%, and 98.64%, respectively, which are significantly better than the values obtained for the existing methods. The objective of this work is to improve classification accuracy and reduce false negatives while keeping computational time to a minimum.

In recent decades, Cellular Networks (CN) have been used broadly in communication technologies. The most critical challenge in the CN was congestion control due to the distributed mobile environment. Some approaches, like mobile edge computing, congesting controlling systems, machine learning, and heuristic models, have failed to prevent congestion in CN. The reason for this problem is the lack of continuous monitoring function at every time interval. So, in this present study, a novel Golden Eagle-based Primal–dual Congestion Management (GEbPDCM) has been developed for the Long-Term Evolution (LTE) Ad hoc On-demand Vector (AODV) network. Here, the Golden Eagle function features will afford the continuous monitoring function to monitor data congestion. Hence, the main objective of this research is to improve the Quality of service (QoS) by optimizing congestion controls. Here, the QoS is measured by different metrics, such as delay, packet delivery ratio (PDR), throughput, packet loss, and energy consumption. Initially, the nodes were created in the MATLAB environment, and the GEbPDCM was activated to predict the data load and estimate the node density to measure the node status. Then, the high data overload was migrated to another free status node to control congestion. Finally, the proposed model efficiency was measured regarding delay, packet delivery ratio (PDR), throughput, packet loss, and energy consumption. The proposed model has scored high throughput at 97.1 Mbps and 97.1 PDR, reducing delay to 67.4 ms and 50.6 mJ energy consumption. Hence, the present model is suitable for the LTE network.

Most augmented reality (AR) pipelines typically involve the computation of the camera’s pose in each frame, followed by the 2D projection of virtual objects. The camera pose estimation is commonly implemented as SLAM (Simultaneous Localisation and Mapping) algorithm. However, SLAM systems are often limited to scenarios where the camera intrinsics remain fixed or are known in all frames. This paper presents an initial effort to circumvent the pose estimation stage altogether and directly computes 2D projections using epipolar constraints. To achieve this, we initially calculate the fundamental matrices between the keyframes and each new frame. The 2D locations of objects can then be triangulated by finding the intersection of epipolar lines in the new frame. We propose a robust algorithm that can handle situations where some of the fundamental matrices are entirely erroneous. Most notably, we introduce a depth-buffering algorithm that relies solely on the fundamental matrices, eliminating the need to compute 3D point locations in the target view. By utilizing fundamental matrices, our method remains effective even when all intrinsic camera parameters vary over time. Notably, our proposed approach achieved sufficient accuracy, even with more degrees of freedom in the solution space.

The molecular subtype of breast cancer plays an important role in the prognosis of patients and guides physicians to develop scientific therapeutic regimes. In clinical practice, physicians classify molecular subtypes of breast cancer with immunohistochemistry(IHC) technology, which requires a long cycle for diagnosis, resulting in a delay in effective treatment of patients with breast cancer. To improve the diagnostic rate, we proposed a machine learning method that predicted molecular subtypes of breast cancer from H&E-stained histopathological images. Although some molecular subtype prediction methods have been suggested, they are noisy and lack clinical evidence. To address these issues, we introduced a patch filter-based molecular subtype prediction (PFMSP) method using spatial transcriptomics data, training a patch filter with spatial transcriptomics data first, and then the trained filter was used to select valuable patches for molecular subtype prediction in other H&E-stained histopathological images. These valuable patches contained one or more genes expressed of ESR1, ESR2, PGR, and ERBB2. We evaluated the performance of our method on the spatial transcriptomics(ST) dataset and the TCGA-BRCA dataset, and the patches filtered by the patch filter achieved accuracies of 80% and 73.91% in predicting molecular subtypes on the ST and TCGA-BRCA datasets, respectively. Experimental results showed that the use of the trained patch filter to filter patches was beneficial to improving precision in predicting molecular subtypes of breast cancer.

Detecting breast cancer through histopathological images is time-consuming due to their volume and complexity. Speeding up early detection is crucial for timely medical intervention. Accurately classifying microarray data faces challenges from its dimensionality and noise. Researchers use gene selection techniques to address this issue. Additional techniques like pre-processing, ensemble, and normalization procedures aim to improve image quality. These can also impact classification approaches, helping resolve overfitting and data balance issues. A more sophisticated version could potentially boost classification accuracy while reducing overfitting. Recent technological advances have driven automated breast cancer diagnosis. This research introduces a novel method using Salp Swarm Optimization (SSO) and Support Vector Machines (SVMs) for gene selection and breast tumor classification. The process involves two stages: mRMR preselects genes based on their relevance and distinctiveness, followed by SSO-integrated WSVM for classification. WSVM, aided by SSO, trims redundant genes and assigns weights, enhancing gene significance. SSO also fine-tunes kernel parameters based on gene weights. Experimental results showcase the effectiveness of the mRMR-SSO-WSVM method, achieving high accuracy, precision, recall, and F1-score on breast gene expression datasets. Specifically, our approach achieved an accuracy of 99.62%, precision of 100%, recall of 100%, and an F1-score of 99.10%. Comparative analysis with existing methods demonstrates the superiority of our approach, with a 4% improvement in accuracy and a 3.5% increase in F1-score over traditional SVM-based methods. In conclusion, this study demonstrates the potential of the proposed mRMR-SSO-WSVM methodology in advancing breast cancer classification, offering significant improvements in performance metrics and effectively addressing challenges such as overfitting and data imbalance.

Nowadays, the pipeline system has the safest, most economical, and most efficient means of transporting petroleum products and other chemical fluids. But, the faults in pipelines cause resource wastage and environmental pollution. Most of the existing works focused either on the surface Crack Detection (CD) or Leakage Detection (LD) of pipes with limited features. Hence, efficient crack detection and leakage monitoring are proposed based on the Acoustic Emission (AE) signal and AE image features using a new Berout Retina Net (BRetN) and Tent Chaotic Kaiming-centric Long Short Term Memory (TCK-LSTM) methodologies. The process initiates from the gathering of input data, followed by preprocessing. Then, the cracks are detected by utilizing Berout Retina Net (BRetN), and the features of AE signals are retrieved. On the other hand, the AE signal is transformed into an AE image using Continuous Wavelet Transform (CWT). Further, the AE image features are extracted, followed by the integration of both the AE signal and AE image features. Further, the optimal features are chosen by using Gorilla Troops Optimizer (GTO). Eventually, the TCK-LSTM model is used for detecting the leakage level of the pipeline. The experimental outcomes illustrated that the proposed framework detected crack and leakage levels with 98.14% accuracy, 95.37% precision, and 98.84% specificity when analogizing over the existing techniques.

Detecting deception in videos has been a challenging task, especially in real world situations. In this study, we extracted the facial action units from the micro-expression, and then calculated the frequency and the number of occurrences of each action unit. To get more information on different scales, we proposed a combination scheme of Multi-Scale Feature (MSF) model and Multi-Head Attention (MHA). The MSF model consists of two CNN with different convolution kernels and GELU is used as the active function. The MHA model was designed to divide the input features into different subspaces and generate attention for each subspace to make the features more effective. We evaluated our proposed method on the Real-life Trial dataset and achieved an accuracy of 87.81%. The results show that the MSF and MHA model could increase the accuracy of deception detection task. And the comparative experiment demonstrates the effectiveness of our proposed method.