基于非负邻域嵌入和非局部正则化的单帧图像超分辨率重建算法

Single-frame Image Super-resolution Reconstruction Algorithm Based on Nonnegative Neighbor Embedding and Non-local Means Regularization

单帧图像超分辨率重建是指利用一幅低分辨率图像,通过相应的算法来获取一幅高分辨率图像的技术。提出了一种基于非负邻域嵌入和非局部正则化的单帧图像超分辨率重建算法,以弥补传统邻域嵌入算法的不足。在训练阶段,首先对低分辨率图像预放大2倍,以保证在放大倍数较大时,高、低分辨率图像块之间的邻域关系也能得到较好的保持;在重建阶段,使用非负邻域嵌入来有效地解决近邻数的选取问题;最后利用图像块的非局部相似性构造非局部正则项对重建结果进行修正。实验结果表明,相对于传统算法,本方法的重建结果纹理丰富、边缘清晰。

Single-frame image super-resolution（SR） reconstruction aims to obtain a high-resolution （HR） image from a low-resolution （LR） input image. To overcome the limitations of traditional neighbor-embedding-based algorithm, we proposed a single-frame image super-resolution reconstruction algorithm based on nonnegative neighbor embedding and non-local means regularization. In the training phase, the LR images are magnified 2 times at first, leading to better pres- ervation of neighborhood between LR and HR images in case of high magnification factor. In the reconstruction phase, non-negative neighbor embedding is employed to select neighborhood number effectively. Finally, a non-local means reg- ularization term is introduced into the final reconstruction process by taking advantage of the non-local similarity be- tween natural image patches. Experimental results demonstrate that the proposed method can achieve results with richer textures and sharper edges compared with those from traditional methods.