基于单一影像局部回归模型修复的Landsat 7 ETM SLC-OFF图像质量评价

Image Quality Evaluation of Landsat 7 ETM SLC-OFF Based on a Single Image Local Regression Model Retrieved

以时段相近的Landsat 5TM影像为参照,通过定性评价(目视)和定量评价(均方根误差、平均差异)方法,分别对单一影像自适应局部回归和单一影像固定窗口局部回归模型修复后的Landsat 7ETM SLC-OFF数据的图像质量进行评价。结果表明,无论是以Landsat 5TM影像还是Landsat 7ETM SLC-OFF影像作为填充影像,得到的修复图像的B1、B2、B3波段均存有一定的条纹,使其应用受到一定限制;但B4、B5、B7波段则没有条纹,这3个波段基本可以应用。对比发现RGF模型方法修复后的影像效果优于FGF模型方法的效果;从填充图像类型看,以Landsat 7ETM SLC-OFF为填充图像修复的影像比以Landsat 5TM为填充图像修复的影像效果更好。

Taking Landsat 5 TM image of adjacent period as the reference, through the qualitative evaluation （subjective visual） and quantitative evaluation （Root-Mean-Square Error, RMSE; Average Difference, AD） methods, the retrieved Landsat 7 ETM SLC-OFF data image quality were evaluated by using a single image adaptive local regression and a single image fixed window local regression model respectively. The results indicated that whether the Landsat 5 TM image or a Landsat 7 ETM SLC-OFF image as filling image, there were some gaps in B1, B2 and B3 band images in filled images, but not in 134,135 and B7 band images. Therefore,B1 ,B2 and 133 band images will be restricted certainly in practice application,and B4,B5 and B7 band images can be applied. The statistical comparison showed that RGF model method has a better effect in filled image than FGF model meth- od does. Viewing from filling image types, Landsat 7 ETM SLC-OFF as the filling image to repair the gaps is better than Land sat 5 TM as the filling image to do that.