基于背景位平面向低位位移的ROI压缩算法研究

针对JPEG2000中一般位移法消耗大量的比特数编码形状信息和最大位移法不能控制感兴趣区域(ROI)与背景(BG)区域相对质量的问题,以及经典多级树集合分裂(SPIHT)压缩算法中忽略小波子带兄弟间相关性的缺点,提出了一种应用于ROI压缩编码的BG位平面向低位位移的移位方法,并对SPIHT压缩编码算法的零树结构进行了改进,在解码端即使BG部分未被反向平移,仍能有较好的结果。通过缩小BG幅值及改进零树结构的SPIHT压缩算法的仿真实验表明,该方法实用有效,适用于较大压缩比及低码率传输的情况。

The research of the ROI compression algorithm based on the shift of background toward the lower bit-plane

To solve the shortcomings of JPEG2 000 that a large number of bits for coding the shape information are consumed in the Generic scaling based method,and the relative qualities of the ROI and background can′t be controlled in the MaxShift method,and the problem in SPIHT that the pertinence of the brother nodes of wavelet sub-bands is ignored,a method that a part of the background bit-plane shifts toword a lower bit-plane and the improvement of zero-tree structure coding algorithm based on the SPIHT are used in the ROI coding.At decoding step,even if the background dosen′t shift contrarily,it still has a better result.The experimental results show that this algorithm is practical,effective and applicable to the cases with greater compression ratio and lower bit transmission rate.