消除影像熱像素雜訊之選擇性向量指向濾波器

Abstract

熱影像雜訊成為數位相機系統的困擾已有多年。人們看到所拍到的影像，若被熱影像雜訊所污染，視覺上感到不舒服，因為這種雜訊破壞整個照片的逼真與美感。然而，數位相機已漸漸成為每個家庭的生活必需品，因此發展熱像素雜訊的移除技術的方法，是必須且重要的。本論文提出消除影像中熱像素雜訊之選擇性向量指向濾波器。我們建構的選擇性濾波器，是以一般向量指向濾波器的概念為基礎。一般向量指向濾波器雖然比其他常見的濾波器耗時，卻在多維影像處理裡，因為有優異的成果而聞名。由實驗結果證明，我們所提出的選擇性向量指向濾波器，比一般向量指向濾波器節省時間，而且在一些測量法裡有較好的成果。雜訊偵測並過濾的正確比率，在實驗結果裡呈現出來，由此可以明顯看出，我們所提出的濾波器，比一般向量指向濾波器的性能好；甚至，以加權式PSNR(WPSNR)值衡量，也跟一些其他濾波器有相近的結果。

Hot spot like noise has been an annoying problem for years in the filed of digital camera system. The contaminated images degraded by this kind of noise do not delight people in human’s visual perception since the contaminations destroy the sense of authenticity and beauty over the images people shoot. Because the digital cameras have almost become the necessities for our everyday life, hence, it is necessary and important to develop a technique to remove hot spot noise in a digital camera. The thesis introduces decision-based vector directional filter to reduce the digital camera corruption incurred by hot spot noise. The decision-based filter we construct is based on the concept of generalized vector directional filter, which is well-known by its outstanding performance in multi-channel image processing although it takes much more time than the other conventional filters. By the simulation results, the proposed decision-based vector directional filter has demonstrated its lower time consumption and higher performance in some measure in comparing to generalized vector directional filter. From the true positive and false positive of the noise reduction obtained in the simulation, it is obviously that the performance of our proposed filter excels that of the generalized vector directional filter. Furthermore, the performance of our proposed filter is comparable to those of some other filters in terms of weighted peak signal to noise ratio.