Attention-Enhanced Deep Learning Model for Parkinson's Diagnosis
Infrared and Visible Image Fusion using Contrast and Edge-Preserving Filters with Image Statistics
Multilevel Thresholding using K-Point Strategy Improved Convergence Based Whale Optimization Algorithm for Image Segmentation
Animal Detection in Fields using Image Processing
Hybrid Approach for Denoising and Segmentation: N2S with Swin Transformerenhanced U-Net
Malaria Detection using Advanced U-Net Deep Learning Model
Identification of Volcano Hotspots by using Resilient Back Propagation (RBP) Algorithm Via Satellite Images
Data Hiding in Encrypted Compressed Videos for Privacy Information Protection
Improved Video Watermarking using Discrete Cosine Transform
Contrast Enhancement based Brain Tumour MRI Image Segmentation and Detection with Low Power Consumption
Denoising of Images by Wavelets and Contourlets using Bi-Shrink Filter
This study presents an AI-based system for early detection of Parkinson's disease using deep learning models Inception V3 and Xception with Attention Mechanism. The system analyzes hand-drawn spiral images, which serve as biomarkers for Parkinson's symptoms like tremors and micrographia. The proposed model extracts critical features from these images using pre-trained convolutional neural networks (CNNs) enhanced with attention layers, ensuring effective classification. The dataset includes spiral drawings from both healthy individuals and Parkinson's patients, allowing the model to learn distinguishing features. The Inception V3 model achieved 100% accuracy, while the Xception model attained 88% accuracy in Parkinson's detection. To evaluate the model's performance, graphs of accuracy against epochs and loss against epochs were plotted to track learning trends. A confusion matrix was generated to analyze misclassifications, and a classification report provided insights into precision, recall, and F1-score. A comparative bar chart was also used to highlight the performance difference between Inception V3 and Xception models. This AI-driven approach provides a non-invasive, cost-effective, and automated diagnostic tool, improving early diagnosis and assisting healthcare professionals in timely intervention.
Infrared (IR) and visible image fusion is a crucial technique in data fusion and image processing. It allows for the accurate integration of thermal radiation and texture details from source images. However, current methods frequently overlook the challenge of high-contrast fusion, resulting in suboptimal performance when replacing thermal radiation target information in IR images with high-contrast information from visible images. To overcome this limitation, a contrast- balanced framework for IR and visible image fusion has been developed. The innovative approach includes a contrast balance strategy for processing visible images, reducing energy while compensating for overexposed areas in detail. Additionally, a contrast-preserving guided filter decomposes the image into energy-detail layers to filter high contrast and information effectively. To extract active information from the detail layer and brightness information from the energy layer, an image statistics technique and a Gaussian distribution of image entropy schemes are introduced for fusing the detail and energy layers. The final fused result is achieved by combining these layers. The final fused result is achieved by combining the detail and energy layers. Comprehensive experiments demonstrate that the proposed method effectively reduces contrast issues while preserving fine details. Additionally, the proposed approach outperformed leading techniques in both qualitative and quantitative evaluations.
The current study presents an innovative multilevel image segmentation method utilizing an improved Whale Optimization Algorithm (WOA). While WOA has shown promise in various optimization tasks, its performance can be limited by a tendency to be trapped in local optima. To address this challenge, the K-point Strategy Improved Convergence WOA (KSICWOA), which enhances optimization efficiency by incorporating a nonlinear convergence factor, an adaptive weight coefficient, and a k-point initialization strategy. The proposed KSICWOA is then applied alongside Otsu's cross variance and Kapur's entropy as objective functions to determine optimal thresholds for multilevel grayscale image segmentation. Experimental results on benchmark functions as well as real-time images demonstrate that KSICWOA surpasses conventional optimization techniques in terms of search accuracy and convergence speed while effectively avoiding local optima. It provides an average improvement of 28.3%, 25.61%, and 7.1% in terms of PSNR, SSIM, and FSIM over the WOA method. Additionally, tests conducted on standard image segmentation datasets confirm that the KSICWOA-Kapur method accurately and efficiently identifies multilevel thresholds.
One of the primary requirements for sustaining a livelihood is agriculture. Low crop productivity is one of the issues facing farmers in the country. Crops destroyed by wild creatures is a major issue in low productivity. The agrarian fields must be defended from any undesirable interruption from creatures. In traditional styles, growers use crackers, electrical walls, direct observation, etc., to keep creatures away from their fields, but it is a threat factor that harms both humans and creatures. The presence of creatures is detected using Image Processing and Machine Learning in the proposed system. The damage to crops caused by wild creatures is dramatically increasing in India. It frequently poses pitfalls to humans and creatures. As wild creatures continue to cause increasing damage to human settlements, tolerance has become difficult. Therefore, an effective solution has been developed to address this situation. With that background, the ideal of this study is to descry wild creatures before entering into the crop fields and enforcing applicable dread- down mechanisms in real time. This paper presents an overview of the methodologies employed in this prototype model, including image segmentation, point birth, and bracket ways. Overall, this study highlights the significance of image processing technologies in advancing the understanding of these models and promoting sustainable relations between humans and wildlife.
Accurate segmentation in medical imaging, particularly for modalities such as Chest X-rays, CT scans, and microscopic images, is critical for diagnosis and treatment. However, noisy and low-quality data can significantly affect performance. This paper presents a novel framework that integrates Noise2Split denoising with a Hybrid Swin Transformer U-Net to enhance segmentation accuracy in these challenging medical imaging tasks. By combining Noise2Split's effective noise reduction with the Swin Transformer's advanced feature extraction and U-Net's robust segmentation architecture, the model efficiently addresses both noise and segmentation challenges. The Swin Transformer effectively captures both local and global context, while the skip connections in U-Net contribute to recovering detailed high- resolution features. Extensive experiments on Chest X-rays, CT scans, and microscopic images demonstrate that this integrated model performs better than traditional methods in terms of segmentation accuracy, making it a valuable tool for clinical applications where imaging quality is compromised.
Malaria continues to affect human lives extensively around the world, requiring urgent medical diagnostic procedures. This paper presents an improved version of the U-Net deep learning method, which identifies malaria within microscopic blood smear images. The segmentation-based feature extraction within U-Net offers superior performance when compared to ordinary deep learning methods, thus leading to better detection results. U-Net delivers precise location detection of diseased areas, which boosts accuracy, while CNN focuses on identification categories, and ANN faces difficulties when identifying complex spatial patterns. The experimental outcomes indicate that U-Net surpasses ANN and CNN approaches by delivering higher values for sensitivity and specificity. The model provides exact detection results and avoids human mistakes while shortening diagnostic time. The system is suited for practical deployment besides offering optimal performance when resources are limited. Data augmentation techniques improve overall generalization properties, which makes the system resistant to different datasets. A modern technological system for automated malaria detection uses deep learning as its foundation.
