Adaptive Segmentation on Extracting Textural and Fractal Patterns for Assessing Mangrove Dynamics Using Multi-spectral Data

This study primarily aims to develop and validate the Kurtosis-based Otsu segmentation technique (KOT), focusing on its effectiveness in separating object boundaries within complex and diverse image environments. Since the fractal dimension measures the complexity and intricateness of an object’s boundary or structure, precise boundary detection ensures the accuracy of the fractal dimension computation. This work elaborates on the investigation of fractal dimensions using Hausdorff approach in order to find self-similar patterns among the diverse mangrove communities of interest present in the individual clusters that make up the target region. The analysis shows that clusters 1, 2, and 3 and cluster 4 have high average Hausdorff fractal dimension values of 0.986 and 0.757, respectively. This suggests the presence of two separate mangrove communities and distinct self-similar patterns within these clusters. The analysis of multi-spectral imagery from the Landsat 8 Operational Land Imager (OLI) is at the heart of the investigation. The study area is located in the Saptamukhi Reserve Forest in the Sundarbans of West Bengal at roughly 21 \(^{\circ }\) 35 \(^{\prime }\) 17.98 \(^{\prime \prime }\) N latitude and 88 \(^{\circ }\) 19 \(^{\prime }\) 09.01 \(^{\prime \prime }\) E longitude. The study uses measures like Mean Squared Error (MSE) and Peak Signal-to-Noise Ratio (PSNR) to assess the segmentation results. After segmenting Band1, the higher PSNR result of 193.31 dB indicates accurate spectral information preservation and decreased distortion. The execution time of 504.66 s and temporal complexity of O(nlogn) of the total KOT segmentation are faster than those of region growing and mean shift segmentation algorithms, pointing to a clear path for efficiently implementing this study for remote sensing-based applications.