Gray Image Compression Using New Hierarchical Self-Organizing Map Technique

This work presents a new hierarchical self-organizing map (NHSOM) to solve image compression problem. NHSOM uses an estimation function to adjust numbers of maps dynamically, and reflects the distribution of data efficiently. Moreover, NHSOM takes splitting LBG to speed up the convergence of SOM, and reduce the training time. Our experimental results show that the proposed NHSOM has good capability in image compression compared with LBG, SOM, HSOM, Modified ART2 and EEMVQ.