基於內插法與梯度的單張影像超解析

Single Image Super Resolution based on Interpolation and Gradient Profile

廖婉婷(Wan-Ting Liao)

指導教授 ： 顏淑惠

淡江大學/工學院/資訊工程學系碩士班/碩士(2015年)

https://doi.org/10.6846/TKU.2015.01038

摘要

本論文提出一個以Reconstruction-based method與Interpolation-based method為基礎的超解析技術，來重建最佳之高解析影像。在Interpolation-based method上我們採用Giachetti et al. [4]作為本篇方法的前處理，並透過大量的統計數據中得知低解析影像與高解析影像之間的Gradient profile存在著規律的變化關係，因此我們利用此特性找出低解析影像與高解析影像之關係式來重建。在最後的實驗中，也驗證了我們的方法能使影像能呈現更多細節。

關鍵字

超解析 ； 內插法 ； 重建式 ； 梯度輪廓

並列摘要

This paper proposes a framework to re-build the high-resolution (HR) images from low-resolution (LR) images by Interpolation-based method and Reconstruction-based method. In pre-process, we adapt Giachetti et al. [4]’s result which base on Interpolation-based method to be our initial image. And through a large of natural image, it can be observe that there has a stable relationship between LR image’s gradient profile and HR image’s gradient profile. Therefore we can apply the relationship to re-establish the HR image. In the experiment, we show that the results are sharp and ringing or zigzagging artifacts are much reduced.

並列關鍵字

Super Resolution (SR) ； Interpolation-based method ； Reconstruction-based method ； Gradient Profile

參考文獻

[1] Xin Li, and Michael T. Orchard. New Edge-Directed Interpolation. In IEEE Trans. on Image Processing, vol. 10, pp. 1521C1527. Oct. 2001.

[2] Andrea Giachetti and Nicola Asuni. Accuracy Improvements and Artifacts Removal in Edge based Image Interpolation. In Proc. 3rd Int. Conf. Computer Vision Theory and Applications (VISAPP08), 2008.

[3] Andrea Giachetti and Nicola Asuni. Fast Artifact-Free Image Interpolation. BMVC, pp. 13.1-13.10, 2008.

[4] Andrea Giachetti and Nicola Asuni. Real Time Artifact-Free Image Upscaling. IEEE TIP, vol. 20, pp. 2760-2768, Oct. 2011.

[6] Jianchao Yang, Zhe Lin, and Scott Cohen. Fast Image Super-resolution based on In-place Example Regression. In IEEE International Conference on Computer Vision (ICCV), pp. 1059-1066, 2013.