基于非局部均值和线性映射的图像超分辨率

Image super-resolution via none-local means learning and linear mapping

针对传统的基于线性映射的超分辨率算法存在丢失一些有价值的图像信息的缺点,提出一种基于非局部均值(NLM)和加权线性映射的图像超分辨率算法。在训练阶段利用迭代反向投影加强双三次插值后的图像;在图像重建阶段,利用加入梯度因素的NLM算法加强双三次插值后的图像并提取特征;对每一个低分辨率块特征,筛选若干个最优的映射函数加权映射得到图像高频信息。该算法保留了较多有价值的图像信息与映射信息,得到了较好的高分辨率图像。实验结果表明,该算法优于比较算法。

Aiming at the problems that traditional linear mapping based image super-resolution methods may lose some valuable image information, the image super-resolution algorithm based on none-local means (NLM) and weighted linear mapping was proposed. In the training phase, iterative back projection was used to strengthen the image after Bi-cubic amplification. In image reconstruction phase, NLM algorithm with added gradient factors was used to strengthen the image after bi-cubic amplification, and then image features were extracted. For features of each low resolution block, a number of optimal mapping functions with weighting coefficients were selected to obtain the corresponding high frequency (HF) details. The proposed method keeps more valuable image information and mapping information, leading to better quality of high-resolution images. Experimental results verify the effectiveness of the proposed method, compared with other state-of-the-art methods mentioned in the paper.