基于兴趣区的编码方法探讨

A Study of ROI-based coding

目的 探讨JPEG2 0 0 0标准中基于兴趣区的编码方法 ,重点讨论了标准中的最大上移方法。方法 最大上移方法缩放兴趣区的系数 ,使得属于兴趣区的所有信息在码流中放在背景区的信息之前。改变图像的量化方式和码流截断点的位置 ,就能获得不同质量的兴趣区和背景区。结果 在低比特率 (大约小于 0 .0 5bpp)时 ,兴趣区和背景区具有近似的图像质量。随着比特率的增加 ,兴趣区的质量迅速提高。结论 最大上移方法处理任意形状的兴趣区时 ,不需要编码形状 ,解码器既不用解码形状 ,也不需要生成兴趣区掩膜 ,因此更加简单。

Objective:To study the encoding methods based on ROI in JPEG2000 still image coding standard and pose emphasis on the Maxshift method. Methods: The Maxshift method was based on scaling of the ROI coefficients so that all the information pertaining to the ROI was placed before the information pertaining to the BG.By varying the quantization of the image and by truncation of the codestream,different quality for the ROI and for the background could be obtained.Results: ROI and BG had very similar image qualities at low bitrates(approximately lower than 0.05 bpp),but with the increase of the bitrates the ROI quality improved much faster. What is more,the quality of the ROI was significantly better than that of the image encoded without an ROI,particularly at low bitrates.Conclusion: the Maxshift method is more simple without the need of shape coding at encoder and without decoding shape and ROI mask generation at the decoder when dealing with arbitrarily shaped regions of interest.