Volume 9, Issue 6 - IJIRTM
November - December (2025)
Impact Factor: 5.86 | Volume 10 | Issue 1
Image Enhancement Using Local and Global Features Optimization
👥 Dr.Rizwana Parveen
📙 Abstract : Image enhancement involves techniques employed to improve the digital quality of images, particularly those of poor quality. One specific method discussed is contrast enhancement, which aims to produce clearer images by redistributing brightness intensity values. This enhancement process utilizes feature optimization techniques, notably particle swarm optimization, to effectively select both local and global features of an image. The optimized features are then applied across the entire image, resulting in improved quality. Experimental results indicate that the proposed threshold-based enhancement algorithm for mixed digital images performs better than traditional methods, especially in cases with heavily damaged pixels, demonstrating its effectiveness in complex scenarios.
🔖 Keywords :️ Image features, PSO, Histogram, Brightness, Contrast.
Sustaining Digital India at the Grassroots: An Empirical Analysis of Common Service Centres and Rural Digital Entrepreneurship
👥 Ankit Singh Bisen, Dr.D. D. Bedia
📙 Abstract : Common Service Centres (CSCs) operate as front-line delivery points for e-governance and digital services in rural India. As the Digital India program matures, questions about the long-term viability of CSCs—especially the business model sustained by Village Level Entrepreneurs (VLEs)—have become pivotal. This study empirically examines the determinants of CSC profitability and sustainability using primary survey data from 200 VLEs across five districts of Madhya Pradesh (Bhopal, Indore, Gwalior, Ujjain, Dewas). Guided by the Resource-Based View and Institutional Theory, we test how Entrepreneurial Orientation (EO), Institutional Support (IS), Digital Skills (DS), Operational Challenges (CH), and Customer Satisfaction (CS) influence Profitability (PI) and Sustainability (SI). Data were analyzed using SPSS v26: descriptive statistics, reliability analysis, Pearson correlation, principal component analysis (PCA), multiple linear regression, and one-way ANOVA. Results show that EO, IS, DS and CS positively and significantly affect SI, whereas CH negatively impacts sustainability and profitability. The regression model explains approximately 71.7% of the variance in sustainability (Adjusted R² ≈ 0.705). Customer satisfaction emerged as the strongest predictor (standardized β ≈ 0.30), followed by institutional support and entrepreneurial orientation. ANOVA indicates significant district-wise differences in profitability (p < 0.05), with urbanized districts outperforming rural ones. We discuss theoretical implications for extending RBV to hybrid public–private digital enterprises, and practical implications for targeted capacity building, infrastructure investment, and policy design. Recommendations include structured training, performance-linked incentives, and localized infrastructure interventions to ensure CSCs function as sustainable rural entrepreneurship hubs.
🔖 Keywords :️ Common Service Centre, Village Level Entrepreneur, Entrepreneurial Orientation, Institutional Support, Digital Skills, Sustainability, India.
Loss-Aware Residual U-Net for Multimodal Brain Tumor Detection and Segmentation
👥 Pramod Mandekar, Chetan Agrawal, Pramila Lovanshi
📙 Abstract : Accurate brain tumor detection and segmentation from multimodal magnetic resonance imaging (MRI) remain challenging due to heterogeneous tumor appearance, modality-specific variations, and severe class imbalance between tumor and healthy tissues. To address these challenges, this paper presents a Loss-Aware Residual U-Net (LA-ResUNet) framework for multimodal brain tumor detection, leveraging complementary information from T1, T1c, T2, and FLAIR MRI modalities. The proposed architecture incorporates residual learning within an encoder–decoder U-Net structure to improve feature propagation and training stability, while a loss-aware optimization strategy, combining Dice loss and focal loss, is employed to effectively handle class imbalance and enhance boundary delineation. The proposed model is evaluated on the benchmark BraTS dataset using standard evaluation metrics. Experimental results demonstrate that the proposed approach achieves a Dice Similarity Coefficient (DSC) of 0.91, sensitivity of 0.93, and overall segmentation accuracy of 98.2%, outperforming conventional U-Net, ResU-Net, and recent multimodal deep learning baselines by a margin of 3–6% in Dice score. In addition, the loss-aware strategy significantly improves the segmentation of tumor core and enhancing tumor regions, reducing false negatives and improving robustness across different tumor sub-regions. The results confirm that integrating multimodal feature fusion with residual learning and loss-aware optimization leads to superior performance in automated brain tumor detection, making the proposed framework a reliable and effective tool for clinical decision support systems.
🔖 Keywords :️ Brain Tumor Detection, Multimodal MRI, Residual U-Net, Loss-Aware Learning, Medical Image Segmentation.