基于结构上下文分类后小面元的后续处理方法

遥感影像在分割、分类处理过程中，不可避免地会出现不合理的小面元，需要进行后续处理剔除。针对当前小面元剔除方法适应性的不足，本文引入结构上下文的处理方法，对复杂小面元进行识别、整理，并基于VC++平台编写了相应的程序，对出现的不合理的小面元进行了剔除处理。实验结果表明，该方法达到了预期目的，取得了较好的效果。

Post-processing Method of Small-area Units Classified Images Based on Structure Context

As small-area units may come forth inevitably during the classification and segmentation of remote sensing images,the post-processing is necessary so as to neaten the classified or segmented images.In this paper,the methodology of using structure coatext to recognize small-area units is introduced elaborately.On the basis of consulting the structure context of image area units,small complicated and unreasonable area units can be recognized and removed by means of the VC++program designed by the authors.The experiment results show that the way of removing small-area units on the basis of structure context can not only recognize and remove disorder area units but acquire a clean and reasonable area image.Also,this method provides the possibility for subsequent process.