常用大气校正模型对图像清晰度提升的对比分析

Comparison of Improving Image Definition with Common-used Atmospheric Correction Models

大气散射、吸收及临近效应等降低了大气调制传递函数而影响遥感图像清晰度,去除大气影响对提高图像清晰度具有重要意义。文章采用典型遥感卫星Landsat-8多光谱数据进行大气校正对图像清晰度提升的研究,基于6S模型、FLAASH模型和黑暗像元法(DOS)模型进行大气校正,得到各谱段地物反射率图像。采用常用的基于图像特征参数(灰度梯度、边缘、熵及频谱)和多光谱图像色彩保真度的清晰度评价方法对校正前后图像清晰度进行评价。结果表明:采用FLAASH、6S和DOS三种模型,大气校正后的清晰度特征参数(以熵为例)较原图平均提升程度分别为27%、10%、1.3%。而色彩保真度方面,各谱段反射率与实际反射率差(以草地为例)的平均值分别为0.018、0.028、0.038。因此,基于辐射传输模型的方法具有更高的大气校正精度,其中FLAASH模型对图像清晰度的提升最明显。

Atmospheric correction is very important to improve image definition, because the scattering, absorbing and adjacent effects of atmosphere can reduce the modulation transfer function and even influence the definition of remote sensing image. In the paper, the multispectral image of the typical satellite Landsat-8 was used in data processing of atmospheric correction. The reflectivity data for each band were obtained using three common atmospheric correction models, i.e. 6 S, FLAASH and DOS. Then the definition of images before and after atmospheric correction was evaluated using four classical definition evaluation methods based on image features（namely, gray-scale gradient, edge, entropy and frequency spectrum） and color fidelity based on ground reflectance. The results indicated that the average definition（evaluated by entropy） had increased by 27%, 10% and 1.3% based on FLAASH, 6 S and DOS respectively. For the color fidelities, the reflectance average differences before and after atmospheric correction, taking the grassland images as an example, were respectively 0.018, 0.028 and 0.038. Therefore, the methods based on radioactive transfer models, especially the FLAASH model, have much better accuracy of atmospheric correction than DOS.