基于分级位平面偏移的多感兴趣区图像编码

Multiple region of interest image coding based on classification bitplane shift

被JPEG2000推荐的Maxshift方式能够高效完成单一感兴趣区(ROI)图像编码,但无法支持不同兴趣度下的多ROI编码。本文提出一种采用分级位平面偏移框架的多ROI编码方式-CBShift。新方法将所有ROI与背景位平面按重要性不同分为重要ROI与背景位平面区(MSRB)、一般重要ROI与背景位平面区(GSRB)及不重要背景区(LSB)。当图像中含有多个ROI时,通过调节每个ROI在GSRB区的位平面偏移量支持不同兴趣度下的多ROI编码。实验显示,新方式不仅具有Maxshift方式的优点,而且支持不同兴趣度下无需ROI形状信息的多ROI编码。

Maxshift method proposed by JPEG2000 can complete single Regions of Interest （ROI） image coding efficiently. However, Maxshift can not support multiple ROI coding based on different degrees of interest. In this paper, a new ROI coding method called CBshift is presented using classification bitplane shift scheme. According to different importance of ROI and Background （BG） bitplane, CBShift divides all bitplane into three parts： MSRB, GSRB and LSB. When multiple ROI is coded in an image, the new method can support multiple ROI coding based on the degrees of interest by adjusting bitplane scaling value of every ROI in GSRB. The experiments show the new method not only has the advantages of the Maxshift method, but also supports multiple ROI coding in different degrees of interest without any ROI shape information.