利用地形正规化校正反立体现象

影像目视判读常会遇到山脊与沟谷的凹凸感与现实相反的反立体现象。消除反立体现象,能有效提高非专业人员对遥感影像的正确使用。立足于反立体现象的成因,本文采用地形正规化模型来校正影像的反立体现象,推导出Lambertian、Cosine-Civco、c校正、b校正这4种地形正规化模型的反立体校正式;对这4种地形正规化模型的反立体校正效果进行了对比,并且与其他5种校正法也进行了对比。通过3个实验区的校正发现,这4种地形正规化模型均能校正反立体现象,但校正影像存在色调偏差;Lambertian、Cosine-Civco的反立体校正影像立体感较强,但影像色调改变较大,视觉效果偏差;c校正、b校正的校正影像在视觉效果和定量指标上都比较接近,基本保持地物光谱信息,校正效果相对较好。从定量指标来看,b校正的反立体校正影像的各指标值整体最小,一定程度代表b校正能取得相对较好的反立体校正效果。与其他方法的对比表明,c校正和b校正的反立体校正不局限于波段个数,在有效消除反立体现象的同时,能相对较好的保留地物光谱信息,有利于影像的定量应用。

Correction of the false topographic perception phenomencn based on topographic normalization

False Top Ographic Perception Phenomena(FTPP)are visual obstacles in applications of remote sensing images. Thid study aims to correct FTPP for non-professionals who use these images. After the causes of FTPP were analyzed, a method was proposed for topographic normalization models to correct FTPP in remote sensing images. Four FTPP correction formula were developed based on the Lambert an model, Cosine-Civco Model, c correction, and b correction; these equations were then compared in three test areas. In addition, the FTPP correction results were compared among the four topographic normalization models and the five existing methods. Results showed that the four topographic normalization models could successfully remove FTPP in images, but introduced some spectral distortions. The FTPP correction results of the Lambert and Cosine-Civco model showed good stereoscopic effects but inferior color balance and hues. The c and b correction produced similar images with good visual effects, while retaining spectral information of ground features. Overall, the quantitative index showed that b correction was slightly better than the three other models for FTPP correction. Comparison with the five existing methods indicated that c and b correction were superior. In conclusion, the c and b correction can effectively remove FTPP in images, regardless of the number of bands involved, there by preserve the spectral information of the images for further quantitative analysis.