International Journal of Image, Graphics and Signal Processing (IJIGSP)

IJIGSP Vol. 17, No. 4, Aug. 2025

Cover page and Table of Contents: PDF (size: 671KB)

Table Of Contents

REGULAR PAPERS

Enhanced Fault Identification in Solar Panels through Binary Cascaded Convolutional Classifiers with Thermal-Visual Image Augmentation

By Sujata P. Pathak Sonali A. Patil

DOI: https://doi.org/10.5815/ijigsp.2025.04.01, Pub. Date: 8 Aug. 2025

Solar power stands as a pivotal renewable energy source for the twenty-first century. However, the optimal functioning of solar panels is often hindered by various faults, necessitating accurate and early defect detection to maximize energy production. Existing solar panel fault identification models encounter challenges such as low precision, difficulty in distinguishing fault types, and poor generalization due to limited and unbalanced data samples. This paper introduces a novel and effective approach, leveraging a Binary Cascaded Convolutional Classifier augmented with visual and thermal image combinations to address these limitations. The proposed model adeptly classifies five distinct types of solar panel faults, including single cell hotspots, diode hotspots, dust/ shadow hotspots, multicell hotspots, and Potential-Induced Degradation (PID) hotspots. Through image augmentation techniques like rotation, shifting, sheering, resizing, jittering, and blurring applied to visual and thermal images, inter-class feature variance is increased. Binary Cascaded Convolutional Neural Network (BCCNN) classifiers are trained using an enriched dataset, each specifically designed to differentiate between dust/ shadow hotspots and other fault categories. The adoption of a binary method significantly enhances precision, allowing for focused fault identification and classification. The proposed model surpasses existing literature in terms of precision (99.8%), accuracy (98.5%) and recall (98.4%), underscoring its effectiveness across all five fault classes. In summary, this research marks a substantial advancement in the realm of solar panel fault identification, presenting a more precise and effective fault detection methodology that has the potential to significantly enhance the maintenance and longevity of solar energy systems.

[...] Read more.
Computed Tomography Image Segmentation Technology Based on ResNet Network Integrated into the Probabilistic Model

By Zhengbing Hu Kostiantyn Zvieriev Oksana Shkurat Andrii Dychka

DOI: https://doi.org/10.5815/ijigsp.2025.04.02, Pub. Date: 8 Aug. 2025

Medical image segmentation is a significant and complex challenge in medical imaging. In recent years, deep learning models have been applied to image segmentation and have shown exceptional performance. However, medical image segmentation has a scarcity of expert-labeled data compared to other deep learning research fields. Therefore, augmenting medical expert-labeled data are primarily the easiest and fastest way to improve the deep learning model’s performance. In this paper, computed tomography image segmentation technology based on the ResNet network integrated into the probabilistic model has been proposed. The proposed segmentation technology is based on the deep learning model of the ResNet50 architecture to extract features from images and initially detect objects of interest and on the probabilistic model with weighted parameters that employs conditional random fields, the GrabCut algorithm, and the argmax function to perform the final detection of objects of interest.
To train, test, and evaluate the effectiveness of the proposed method, appropriate chest CT datasets were identified to solve the task of segmenting the lung cavity, the liver and areas affected by COVID-19. The proposed image segmentation technology demonstrates segmentation accuracy results of 73.12% by Dice Score for the COVID-19 disease dataset, 97.71% for the lung cavity dataset, and 98.36% for the liver dataset, which perform better than state-of-the-art solutions.
The proposed image segmentation technology has been compared with state-of-the-art technologies (SegNet, UNet, and FCN-ResNet50) for CT segmentation to demonstrate the effectiveness of the method. The positive outcomes strongly suggest the significant potential of the proposed image segmentation technology. According to the obtained results, the proposed image segmentation technology can be a useful auxiliary tool for doctors to segment CT images for further analysis and monitoring of statistical and dynamic indicators.

[...] Read more.
An IoMT enabled Deep Insight of MR Images for Brain Tumor Segmentation with Classification Using an Elevated UNet-RESNet Model

By Surendra Kumar Panda Ram Chandra Barik Ganapati Panda Suvamoy Changder

DOI: https://doi.org/10.5815/ijigsp.2025.04.03, Pub. Date: 8 Aug. 2025

Brain tumors are a prominent cause of mortality on a global scale. The American Brain Tumor Association reports 90,000 primary brain tumor diagnoses annually, highlighting the need for improved diagnostic methods. Delaying brain tumor identification can result in significant financial costs and considerable suffering for patients. Timely identification of brain tumors is crucial for preserving both financial resources and human lives. Physicians’s manual identification of brain tumors is quite challenging. Early and precise brain tumor detection is crucial to addressing these concerns. The incorporation of the Internet of Medical Things (IoMT) coupled with deep learning (DL) is essential for advancing contemporary healthcare solutions. The proposed work presents the IoMT-UNet-ResNet model, an advanced DL method designed specifically for accurately identifying and classifying brain tumors in MR image data. By harnessing the potential of the IoMT, the model effortlessly combines UNet for precise spatial delineation and ResNet-50 for sophisticated feature learning, resulting in outstanding accuracy. This model proves to be an invaluable asset for radiologists, as it simplifies and improves the precision of brain tumor analysis through the use of MRI data. The IoMT enables radiologists to effortlessly access and analyze diagnostic information in real-time, leading to enhanced patient care and results in the field of neuroimaging. The proposed IoMT-UNet-ResNet model outperforms by comparing and validating the existing technique.

[...] Read more.
Enhancing In-loop Filter of HEVC with Integrated Residual Encoder-Decoder Network and Convolutional Neural Network

By Vanishree Moji Bharathi Gururaj Mathivanan Murugavelu

DOI: https://doi.org/10.5815/ijigsp.2025.04.04, Pub. Date: 8 Aug. 2025

High Efficiency Video Coding (HEVC) often known as H.265 is a video compression method that outperforms its predecessor H.264. In HEVC, an in-loop filter is an additional processing step that removes compressing artifacts from decoding video frames while improving visual quality. This research article proposes an improved in-loop filter that incorporates a Residual Encoder-Decoder Network based Deblocking Filter (REDNetDF) and a Convolutional Neural Network based Sample Adaptive Offset (CNN-SAO) filter, which together eliminates the smallest range of artifacts in compression video frames. The quantization frame is subjected to REDNetDF, which removes a minute number of blocking artifacts from the compressed frame. To eliminate the ringing artifacts in the compressed frame, CNN-SAO filter is used. The proposed method is used to evaluate the publicly available UVG dataset. To demonstrate efficiency, the new model is evaluated using a variety of metrics. The outcome of this study provides better results like PSNR of 49.7 dB and the SSIM of 0.97 in comparison with other techniques. Besides, the model's outcome indicates an MSE of 1.8 and saves 24.9% more bits on average to provide the same level of quality as previous techniques. The proposed framework also suppresses time complexities regarding encoding and decoding times with the results of 90.5 and 4.5 seconds on average correspondingly.

[...] Read more.
An Effective Hybrid HBA-MAO for Task Scheduling with a Hybrid Fault-Tolerant Approach in Cloud Environment

By Manoj Kumar Malik Hitesh Joshi Abhishek Swaroop

DOI: https://doi.org/10.5815/ijigsp.2025.04.05, Pub. Date: 8 Aug. 2025

"Cloud computing" refers to internet-based computing on demand and describes an incredibly scalable technology used by working-class and non-working individuals globally. Fault-tolerant task scheduling is an essential tool used by end users and cloud suppliers. Finding the best resource for the specified input task presents a key challenge for fault-tolerant task schedulers. The studies that have already been done have attempted to address each of these complex issues independently. Still, it is tricky to optimize resources and provide fault tolerance at the same time. In this paper, an effective hybrid HBA-MAO and hybrid fault-tolerant mechanism in cloud computing are designed to appropriate task scheduling in VMs without delay and failure. Various tasks submitted by users and virtual machines are taken as input for the proposed approach. Hybrid Honey Badger Optimization Algorithm (HBA) and Mexican Axolotl Optimization (MAO) are used in this proposed for priority based optimal task scheduling. These scheduled tasks are assigned to the VM for execution. A fault-tolerant mechanism is immediately carried out if the tasks are not completed successfully. The hybrid reactive and proactive fault-tolerant mechanism is used in this proposed approach for a high level of fault tolerance. The proposed approach attains better performance, like 70 sec of response time, 13% of resource utilization and 95% success rate. This approach uses resources efficiently by reducing resource consumption, so it is the best choice for fault-tolerant aware task scheduling.

[...] Read more.
Spoof-formerNet: The Face Anti Spoofing Identifier with a Two Stage High Resolution Vision Transformer (HR-ViT) Network

By Mudunuru Suneel Tummala Ranga Babu

DOI: https://doi.org/10.5815/ijigsp.2025.04.06, Pub. Date: 8 Aug. 2025

Face anti-spoofing (FAS) detection is essential for assuring the safety and dependability of facial identification systems. This study introduces the implementation of a new approach called Spoof-formerNet, which utilizes the high-resolution vision transformer (HR-ViT) system for detecting face anti-spoofing. The Vision Transformer (ViT) architecture has revealed remarkable execution in numerous computer vision applications, and we are now applying it to the intricate field of spoof detection. In order to distinguish between real faces and spoofing attempts, the Spoof-formerNet is engineered to detect minute details and subtle elements embedded in facial photos. We have conducted experimental research wherein the model is trained independently on color (RGB) and depth data in parallel using two streams of HR-ViT networks. Before applying to a classification head, the features from the two streams were concatenated. Spoof-formerNet is trained and tested using well-known benchmark datasets such as CelebA-Spoof, CASIA-SURF, WMCA, and MSU-MFSD, which are commonly used in the field of anti-face spoofing. The suggested model excels in performance and is cutting-edge in identifying genuine faces from spoofing assaults. We assess the model's efficacy by providing comprehensive findings, such as Area Under the Curve (AUC), Attack Presentation Classification Error Rate (APCER), Bona Fide Presentation Classification Error Rate (BPCER), Equal Error Rate (EER), and Average Classification Error Rate (ACER). The results of this work show how cascaded high-resolution vision transformer networks can be used to improve the safety of facial recognition approaches in real-world applications, in addition to advancing facial anti-spoofing technology. The Spoof-formerNet method for face anti-spoofing detection shows good results, with an average AUC of 99.22 and average APCER, BPCER, and ACER of 0.95, 0.66, and 0.81 correspondingly.

[...] Read more.
Evaluation of Deep Learning Approaches to Detect Choroidal Neovascularization

By Sahil Chukka Vardhanika Jagtap Naveen Patel Sudiksha Jadhav Mimi Cherian Jinesh Melvin Y. I.

DOI: https://doi.org/10.5815/ijigsp.2025.04.07, Pub. Date: 8 Aug. 2025

In ophthalmology, Choroidal Neovascularization (CNV) is a serious medical disease that, if left untreated, frequently results in significant vision loss. In this investigation, we investigate the evaluation and working of deep learning models, notably basic Convolutional Neural Networks (CNN), ResNet18, ResNet50, VGG16, VGG19, Vision Transformers, EfficientNetV2L, MobileNetV2 and InceptionV3 for identification and classification of CNV in Optical Coherence Tomography (OCT) images. The Kermany dataset, which includes OCT images of both CNV-patients and non-CNV patients (Normal OCT images) are utilized for this paper. The dataset was further used in three different versions based on validation and training split. The images from the dataset are already pre-processed and labelled so no pre-processing operations were performed, how- ever resizing of images have been performed according to the models. The deep learning models are trained and evaluated on standard performance metrics such as precision, recall, accuracy, F1-score, etc. All things considered, our work shows the evaluation of deep learning models to classify OCT images that show the presence of CNV. Based on all three dataset versions, the findings of our study confirm that ResNet18, VGGNet19, and MobileNetV2 beat all other approaches and achieved an average accuracy of 1. Additionally, Vision Transformer and Effi- cientNetV2L demonstrated strong performance, averaging 0.99 and 0.96 accuracy on each of the three dataset versions, respectively. These models have the potential to help ophthalmologists detect CNV early and monitor it, which may lead to prompt treatment and better vision preservation for patients.

[...] Read more.
Application of Tensor Networks Analysis to Optimize Traffic Management in a Critical Information and Telecommunications Network

By Oleksandr Lavrut Tetiana Lavrut Victoria Vysotska Zhengbing Hu Yuriy Ushenko Dmytro Uhryn

DOI: https://doi.org/10.5815/ijigsp.2025.04.08, Pub. Date: 8 Aug. 2025

The article investigates the task of optimising traffic management in critical information and telecommunication networks in order to ensure a guaranteed quality of user service, particularly in emergencies. A method of tensor analysis of networks is proposed, using a formalised description of the system in the form of tensors of message lengths, delays and bandwidth of channels. The network is modelled as a simplified complex, and routing is implemented through a tensor equation of connection between network parameters in different coordinate systems. Experimental calculations using examples with dynamically variable topology have shown:

•Reduction of average multipath message delivery latency by 9–40% depending on traffic intensity,
•Probability of packet delivery at or above 0.999 under high loads (200-300 messages/s),
•Zero jitter due to the even distribution of delays between paths,
•The ability to adaptively fragment messages in nodes to reduce latency,
•Increasing the efficiency of resource use compared to single-track models.
 
The use of a tensor apparatus provides stable and scalable routing in an unstable network topology. The method allows you to take into account the heterogeneity of traffic, adapt to the loss of nodes or channels, and maintain guarantees of quality of service in real time. The proposed approach is of practical importance for information and telecommunication systems used in emergencies, in particular for coordinating the actions of emergency rescue services, emergency medicine, civil protection, military units, control of drones and robotic means in the face of infrastructure loss. Potential stakeholders include state and municipal security services, operators of critical networks (energy, transport, healthcare), developers of automated control systems, and manufacturers of secure communication equipment. The proposed method can be integrated into decentralised networks with limited resources and variable topology, where traditional routing approaches do not guarantee sufficient quality of service.

[...] Read more.