Novel set partitioning algorithm in image coding

This paper provides an insight into the principle of set partitioning in image coding. The state-of-the-art SPIHT algorithm is questioned to be not optimum with its fixed set-partitioning way. To further explore the energy clustering nature of subband coefficients, the set with extended hierarchical zerotree or group of zerotrees is introduced to represent the actual zero blocks emerging in the partitioning process. Moreover, through simple yet effective prediction, the sets containing extended zerotrees with different shape and size are created, adapting to local distribution of wavelet coefficients of the image. With a similar procedure to SPIHT, a novel algorithm is proposed and tested. The simulation shows that the new algorithm can achieve better performance with lower memory requirement and negligible additional complexity in comparison with existing SPIHT.