Enhanced FCM Algorithm with Hierarchical Segmentation

To address the challenges of long execution time and high computational load in traditional FCM algorithms, we propose an enhanced FCM algorithm. The improved methodology first partitions the input image into window-sized sub-blocks using systematic grid division. For each sub-block, we extract representative feature vectors that capture local image characteristics. These feature vectors then undergo coarse segmentation through standard FCM clustering implementation, where the algorithm minimizes the objective function through iterative centroid updates and membership calculations. Following the coarse segmentation stage, the algorithm specifically targets edge sub-blocks identified through boundary detection. For these critical regions, we perform pixel-level feature extraction using techniques like intensity gradients or texture descriptors. The fine segmentation phase then applies FCM clustering at the pixel level within these edge sub-blocks, ensuring precise boundary delineation. This two-tiered approach reduces computational complexity by limiting pixel-level processing to strategically important areas while maintaining overall segmentation quality through sub-block level processing for homogeneous regions. The implementation leverages matrix operations for efficient feature extraction and utilizes optimized distance calculations in the clustering process. Key functions include sub-block division using sliding window techniques, feature vector normalization, and adaptive thresholding for edge sub-block identification. This hierarchical segmentation framework demonstrates improved runtime performance and enhanced segmentation accuracy compared to conventional FCM approaches, effectively solving the computational bottlenecks associated with traditional implementations.