Beyond Boundaries: Region Filling with Image Data Clustering
K. Hari Nagarubini, K. Suthendran · 2024
Original region filling algorithms for image processing face challenges in computational complexity, limiting their use in real-time and resource-constrained applications. The proposed two-step method priorities efficiency. To streamline clustering, the image preprocessing first removes all-zero rows and columns, focusing the algorithm on informative data for identifying fill regions, without the need for explicit boundary detection. After filling occurs, it reintroduces the initially removed rows and columns to retain the original structure. This temporary zero removal boosts clustering efficiency while preserving data integrity. The proposed algorithm surpasses traditional methods, especially when dealing with large and single hole images, as evidenced by experimental results.