利用色彩濾波矩陣對單張圖片做超解析度圖像放大演算法

Abstract

超解析度(Super-Resolution)一般泛指影像利用單張或是多張的低解析度影像透過重建、學習、補償以及後處理的方法來產生出具有豐富資訊及細節的高品質影像。高解析度的影像在日漸進步的高科技社會中越來越重要，舉凡醫學影像、工業影像、衛星影像甚至生活中監控系統影像及普通生活照都是極度被需求的。低解析度的影像容易造成模糊與方格效應(Lattice Effect)，而超解析度(Super-Resolution)技術可以有效的改善這些問題。 在本論文中，我們提出一個基於重建色彩濾波矩陣(Color Filter Array)的超解析度影像演算法。相較於現有以多張重建或以學習為基礎的超解析度方法，我們的方法是非常有效率的。不僅具有銳利的邊界部分，亦擁有豐富的細節。實驗結果顯示本論文所提出的新方法對於增加影像解析度有令人滿意的效果。

“Super-Resolution” generally refers to the images use a single frame or multiple frames of low resolution to produce high quality images with rich information and details. The high-resolution images play more and more important roles in the high-tech community nowadays. Different kind of high resolution images are extremely demanded such as medical images, industrial images, and satellite images, surveillance camera images and even photos of a daily life. Low-resolution images may cause the “Lattice effect” and blur. However, super-resolution technology can improve these problems effectively. In this paper, we propose a super-resolution algorithm based on color filter array reconstruction. This algorithm only uses a single low-resolution image to scale up. Comparing to some proposed example-based super-resolution algorithms or some algorithm scale images up by multiple frames, our method is much more efficiently. The experiment and results show that the algorithm of this paper in increasing resolution of images has a satisfactory performance.