利用基礎層和細節層的濾波器分解實現彩色影像增強

Abstract

在過去十年，強化彩色影像一直是一個受歡迎的主題，而它的目標為改善原始影像之視覺品質。由於濾波器在影像處理被認為是最重要的運算，因此我們使用濾波器去實現這個應用，特別是能將邊際保留的模糊濾波器，它們對一些應用而言都十分重要。我們使用四種不同的邊際保留模糊濾波器，利用它們得到被模糊但還保有邊際的基礎層。而它們分別為雙向濾波器、嚮導濾波器、使用區域轉換的濾波器和L_0模糊濾波器，利用這些濾波器所產生的基礎層可計算出細節層，我們對基礎層和細節層做處理，以達到彩色影像強化的效果。 一張相片包含很多視覺的資訊，在人類的視覺感知裡，邊際對於神經感測到一個畫面的解釋而言是相當重要的一環。因此利用將基礎層和細節層分開，能區隔一張相片之邊際和細節部分，進行運算時才不會被同步處理和互相影響。 在這篇論文中，我們會一一介紹四種邊際保留模糊濾波器的理論和利用它們實做彩色影像強化，而在最後會對四種濾波器的結果做比較。

In the past decades, color image enhancement has been a popular topic, and whose goal is to improve the visual quality of the original image. We use filters to implement the application because of filtering is arguably the most important operation in image processing. Particularly, edge-preserving smoothing filters are a fundamental building block for several applications. We utilize four kinds of edge-preserving smoothing filters to obtain the base layers, which are blurred but still retain their edges. They are bilateral filter, guided filter, filters based on domain transform, and L_0 smoothing filter. Using base layers to produce detail layers, and enhancing images by processing base layers and detail layers. Photos contain well-structured visual information. In human visual perception, edges are vital for neural interpretation to make the sense of the scene. We decompose base layer and detail layer, and separate edges and details. Therefore, we can process them independently and do not have effect to each other. In this thesis, we introduce the four kinds of edge-preserving smoothing filters and implement color images enhancement based on them. In the end, we compare the experimental results of the four filters.