基于多引导滤波器的单幅图像超分辨率技术

发布时间：2019-08-14 13:41

【摘要】：提出了一种基于多引导滤波器的单幅图像超分辨率方法。首先,该方法通过大量的自然图像建立高低分辨率图像块样本训练库,并通过聚类算法将具有相似性质的高低分辨率样本块进行聚类;其次,将输入低分辨率图像进行重叠分块,并在样本库中搜索最近邻的高低分辨率样本聚类;再次,将输入低分辨率图像块作为输入图像,与样本库中最近邻的低分辨率聚类样本作为引导图像,运用本文提出的多引导滤波器计算引导滤波器的参数;最后,利用样本库中最近邻的高分辨率聚类样本和引导滤波器的参数,通过多引导滤波器就可以重构高分辨率图像。实验结果表明,本文算法不仅能很好地重构图像的高频细节,还能很好地恢复图像的纹理特征。
[Abstract]:A single image super-resolution method based on multi-pilot filter is proposed. Firstly, the high and low resolution image block sample training database is established through a large number of natural images, and the high and low resolution sample blocks with similar properties are clustered by clustering algorithm. Secondly, the input low resolution image is overlapped and the nearest neighbor high and low resolution sample clustering is searched in the sample database. Thirdly, the input low-resolution image block is used as the input image, and the nearest neighbor low-resolution clustering sample in the sample database is used as the guided image, and the parameters of the pilot filter are calculated by using the multi-pilot filter proposed in this paper. Finally, the high-resolution image can be reconstructed by using the parameters of the nearest neighbor high-resolution clustering sample and guiding filter in the sample database. The experimental results show that the proposed algorithm can not only reconstruct the high frequency details of the image, but also restore the texture features of the image.
【作者单位】：西京学院电子信息工程系;西安交通大学电信学院计算机科学与技术系;
【基金】：国家自然科学基金(61473237)
【分类号】：TN713;TP391.41

【相似文献】