星载遥感图像的二维均值预测  

2-D mean prediction for space-borne remote-sensing images

作　　者：魏永旺[1,2,3,4,5] 罗海波[1,3,4] 李德强[1,3,4] 张承宁[1,2] 吴永国[1,2,3,4] 

机构地区：[1]中国科学院沈阳自动化研究所,辽宁沈阳110016 [2]中国科学院研究生院,北京100049 [3]中国科学院光电信息处理重点实验室,辽宁沈阳110016 [4]辽宁省图像理解与视觉计算重点实验室,辽宁沈阳110016 [5]北华大学电气信息工程学院,吉林吉林132021

出　　处：《计算机工程与设计》2011年第6期2061-2064,2091,共5页Computer Engineering and Design

基　　金：中国科学院沈阳自动化所先进制造基地支持基金项目(07F4090401)

摘　　要：根据星载遥感图像的相邻像元间具有平滑性、均一性等特点,提出了一种二维均值预测方法(Mean预测)。该方法以被预测图像的两个相邻像素作为预测基准,通过计算两点灰度值的平均值然后向下取整,得到预测值,Mean预测残差近似服从Laplace分布,方法可逆。实验仿真同JPEG-LS标准中的MED预测、CCSDS预测和Zig-Zag预测做了比较,统计结果表明,Mean预测法与其它方法相比,在相关系数、熵值、均值以及均方差的性能上均有一定程度的提高,计算量小且硬件实现简单,可用于星载遥感图像的实时无损压缩处理。A two-dimensional mean prediction method is proposed, according to characteristics ofspace-borne remote-sensing images, such as smoothness and uniformity among the neighboring pixels. The predicting value is obtained by computing the mean of two neigh-boring pixels, and then be rounded to the nearest integer less than or equal to the mean. The differences of the mean prediction approximately belong to the Laplacian distribution, and this method is reversible. The result of the mean prediction method is compared with MED prediction, CCSDS prediction and Zig-Zag prediction through the simulation experiments. The statistic results show that the performance of the mean prediction method is better than other predictions to some extent in term of the correlation coefficient, entropy, mean and mean Square deviation of difference images. Moreover, it can be easily realized in the hardware system for small computational amount, and be used in real time lossless compression for space-borne remote-sensing images.

关 键 词：星载遥感图像 二维均值预测 MED预测 相关系数 熵 

分 类 号：TP751.1[自动化与计算机技术—检测技术与自动化装置]