基于分类字典学习的遥感图像超分辨率方法

Remote-sensing Image Super-resolution Algorithm Based on Classified Dictionary Learning

传统的超分辨率方法存在图像重构时间长,重构质量有待改进的问题。因此,文章针对遥感图像对传统的超分辨率方法进行了改进。主要利用原始图像的局部二值模式(LBP)纹理特征对图像进行分类识别,学习分类字典,并使用对应类别字典对低分辨率图像进行超分辨率重构。该方法的优势在于既加快了重构速度,又有效改善了重构图像的质量。试验结果证明了该方法相对于传统方法的优越性。

The traditional super-resolution method has problems that the reconstruction time is too long and the reconstruction quality needs to be improved. Therefore, in view of the remote sensing image, this paper improves the traditional method of super-resolution. The method mainly uses the local binary pattern（LBP） texture feature of the original image to classify and recognize them, and uses the corresponding category dic- tionary of low-resolution to reconstruct the super-resolution image. The advantage of the method is speeding up the reconstruction and effectively improving the quality of the reconstruction image. The experimental result shows the superiority of this method compared with the traditional method.