基于零树结构的可伸缩编码方法研究实现

Scalable Image Coding Method Based on Zerotrees of Wavelet Coefficients

文章利用小波变换多分辨分解特点和Ｓｈａｐｉｒｏ提出的ＥＺＷ编码的优点，实现了一种基于零树结构的可伸缩性编码新方法，其编码特点是可以输出分辨率多位率压缩位流，有效地改进了ＥＺＷ方法只能应用于单一分辨率视频服务的不足。该算法首先对小波变换系数进行一致标量量化；其次对量化后不为零的重要低频系数按照位平面次序进行编码；最后根据零树结构对高频小波系数，从低分辨率开始逐频带搜索得到不为零的重要系数，对搜索到的高频重要系数同样按照位平面次序进行编码，直至最高频带小波系数。该算法的特点是具有可伸缩性，即其编码位流不但可以进行多分辨率解码，而且可以进行多位率解码。实验结果表明，该压缩方法性能优于Ｓｈａｐｉｒｏ的ＥＺＷ方法，能在一定程度上提高峰值信噪比（ＰＳＮＲ）。

Based on the proper'ty of wavelet transform and the advantage of Shapiro's EZW coding algorithm,this paper proposes a scalable image coding method using zerotrees of wavelet coefficients. Firstly form a subband pyramid and quantize all wavelet coefficients using uniform scalar quantizer. Secondly encodes the magnitude of nonzero wavelet coefficients at low frequency subband in bit plane order. Finally beginning from coarsest resolution subband of high fre- quency and ending at finest resolution subband of high frequency, independently search for all nonzero wavelet coeffi- cients in each subband using zerotrees and encode them in bit plane order separately. Differing significantly from Shapiro's,this approach endows the bit stream with the multirate and the multiresolution property simultaneously.Experi- ment results show that the coding method is superior to the EZW algorithm in PSNR and visual effect.