A fast single-image super-resolution method implemented with CUDA
Image super-resolution (SR) plays an important role in many areas as it promises to generate high-resolution (HR) images without upgrading image sensors. Many existing SR methods require a large external training set, which would consume a lot of memory. In addition, these methods are usually time-consuming when training model. Moreover, these methods need to retrain model once the magnification factor changes. To overcome these problems, we propose a method, which does not need an external training set by using self-similarity. Firstly, we rotate original low-resolution (LR) image with different angles to expand the training set. Second, multi-scale Difference of Gaussian filters are exploited to obtain multi-view feature maps. Multi-view feature maps could provide an accurate representation of images. Then, feature maps are divided into patches in parallel to build an internal training set. Finally, nonlocal means is applied to each LR patch from original LR image to infer HR patches. In order to accelerate the proposed method by exploiting the computation power of GPU, we implement the proposed method with compute unified device architecture (CUDA). Experimental results validate that the proposed method performs best among the compared methods in both terms of visual perception and objective quantitation. Moreover, the proposed method gets a remarkable speedup after implemented with CUDA.
KeywordsSuper-resolution Self-similarity GPU CUDA
The research in our paper is sponsored by National Natural Science Foundation of China (Nos. 61701327, 61711540303, 61473198), also is supported by the Priority Academic Program Development of Jiangsu Higher Education Institutions (PAPD) Fund, Jiangsu Collaborative Innovation Center on Atmospheric Environment and Equipment Technology (CICAEET) Fund.
- 6.Glasner, D., Bagon, S., Irani, M.: Super-resolution from a single image. In: 2009 IEEE 12th International Conference on Computer Vision, pp. 349–356. IEEE (2009)Google Scholar
- 8.Zeyde, R., Elad, M., Protter, M.: On single image scale-up using sparse-representations. In: International Conference on Curves and Surfaces, pp. 711–730. Springer, Berlin (2010)Google Scholar
- 12.Dong, C., Loy, C.C., He, K., Tang, X.: Learning a deep convolutional network for image super-resolution. In: European Conference on Computer Vision, pp. 184–199. Springer, Cham (2014, September)Google Scholar
- 13.Huang, J.B., Singh, A., Ahuja, N.: Single image super-resolution from transformed self-exemplars. In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp. 5197–5206 (2015)Google Scholar
- 20.Yang, C.Y., Huang, J.B., Yang, M.H.: Exploiting self-similarities for single frame super-resolution. In: Asian Conference on Computer Vision, pp. 497–510. Springer, Berlin (2010)Google Scholar
- 22.Wu, W., Zheng, C.: Single image super-resolution using self-similarity and generalized nonlocal mean. In: TENCON 2013-2013 IEEE Region 10 Conference, pp 1–4. IEEE (2013)Google Scholar
- 23.Sun, K.Z., Li, J.D., Xu, S.Y.: Gpu-Accelerated non-local means super-resolution reconstruction. In: Proceedings of the 3rd International Conference on Multimedia Technology, ICMT, pp. 1242–1249 (2013)Google Scholar
- 25.Document Page of Nvidia. http://docs.nvidia.com/cuda/cuda-c-best-practices-guide. Accessed 18 Dec 2017
- 26.Home Page of Nvidia. https://developer.nvidia.com/. Accessed 18 Dec 2017
- 27.Garcia, V., Debreuve, E., Nielsen, F., Barlaud, M.: K-nearest neighbor search: fast GPU-based implementations and application to high-dimensional feature matching. In: 2010 17th IEEE International Conference on Image Processing (ICIP), pp. 3757–3760. IEEE (2010)Google Scholar