基于张量的单幅图像超分辨算法

Tensor-based super-resolution algorithm for single image

图像的边缘细节信息直接影响图像的视觉质量。为了尽可能地保持图像边缘结构信息,提高超分辨率图像的质量,提出了一种基于张量的单幅图像超分辨算法。该方法利用张量对图像局部几何特征进行描述,然后根据采样点的局部特征估计待插值点的局部特征,最后通过这一估计的特征计算待插值点的灰度值。实验结果表明基于张量的超分辨方法能够较好地保持图像中的边缘结构信息,峰值信噪比(PSNR)、结构相似性系数(SSIM)等客观评价指标和主观视觉效果都比较好。

Edge details of the image directly affect the visual quality of the image. In order to maintain structure information of the image edges as much as possible, and then improve the quality of super-resolution images, a tensor-based single image super-resolution algorithm was proposed. Firstly, the local geometric characteristics of image were described by tensor, then the local characteristics of the interpolation points were estimated according to that of the sampling points. Finally, the gray values of the interpolation points were calculated by the estimated characteristics. The experimental results show that the super-resolution method based on tensor can better preserve the structure information edges in the image, and performace better in Peak Signal-to-Noise Ratio （PSNR）, Structural SIMilarity （SSIM） and subjective visual effect.