提出了一种感兴趣区域的多描述SPIHT编码方法.该方法将具有分层内嵌特性的SPIHT码流编码成多个描述子,通过对所有零树重新进行重要性排序,为每个描述子分配不同编码速率的完全编码码流和冗余码流.在分配速率时,优先对感兴趣区域分配较多的冗余,而对背景区域或非感兴趣区域分配速率较低的冗余,以进一步增强感兴趣区域的抗误码以及抗分组丢弃能力.文中对编码过程中零树、冗余树及其编码速率之间的关系进行了定量分析,并给出了重要性排序和冗余树分配的具体算法.实验证明,该文方法同其它感兴趣区域编码方法相比.可以进一步提高感兴趣区域在分组丢弃情况下的恢复图像质量.
A new multiple description SPtHT coding for region of interest(ROI) is proposed to combat packet loss. The method codes the SPIHT bit stream into multiple descriptions, each with a variable number of fully coded original trees and partially coded redundant trees. After reordering the zero trees of wavelet according to their importance, the method assign different coding rates of original trees and redundant trees to each description. The more important parts of each tree are sent in more description than the less important parts, and in order to further improve the error resilience of ROI, more redundant trees are assigned to ROI than non-ROI. The relationship among zero trees, redundant trees and their coding rates are analyzed, and the algorithm of sorting process and assigning strategy of redundant trees are also presented in detail. Simulation results show that the method can further improve the reconstructed quality of ROI in the case of packet loss compared to other ROI coding methods.