基于广义RoI的遥感图像压缩
详细信息 本馆镜像全文    |  推荐本文 | | 获取馆网全文
摘要
针对已有的基于RoI的图像压缩方法的不足 ,提出在初步分割出目标后 ,进一步在目标的语义方面进行改善 ,不仅保留对视觉有重要作用的边缘信息 ,而且基于图像的内容以及目标的语义 ,保留与目标相关的区域信息 .并把这部分区域与已分割出的目标构成广义RoI .最后 ,对广义的RoI和背景分别用不同的压缩方法进行压缩 .对广义RoI采用无损压缩以保留区域信息 ,对其余部分采用有损压缩以保留边缘信息 .实验证明本文方法允许目标近无损分割 ,并能在不降低压缩比的前提下增强对重建目标的理解
After object was segmented, some post processes were implemented to preserve some regions relative to object. These regions and the segmented object would form the extended RoI. The extended RoI was compressed by using lossless scheme to preserve detail information and the rest was compressed by using lossy method to maintain edge information. It was shown the algorithm enhanced comprehension for reconstructing object and permitted lossless near segmentation.
引文
[1]NisterD ,ChristopoulosC .Losslessregionofinterestcoding.SignalProcessing,1999,78(1):1~17
    [2]AskelofJ ,CarlanderML ,ChristopoulosC .RegionofinterestcodinginJPEG 2000.SignalProcessing:ImageComm.,2002,17(3):105~111
    [3]DuchowskyAT .Representingmultipleregionsofinter estwithwavelets.In:AwardsBed.Proc.SPIEonVisualCommunicationsandImageProcession.SanJose:1998.SanJose:SPIE ,1998.975~986
    [4]王碧泉,陈祖荫.模式识别的理论、方法和应用.北京:地震出版社,1989.

版权所有:© 2023 中国地质图书馆 中国地质调查局地学文献中心