數位音訊放大的處理與影像脈衝型雜訊的濾除

Abstract

隨著多媒體的普及, 每天都有大量的影像和音訊產生並且在個各地方傳遞，人們對影像和音訊品質的要求也越來越高。在音訊方面，由於原始音訊通常是很小的訊號，而必須被放大，然而傳統的類比式放大器時常會有放大失真現象，直到70年代，新型的放大技術，Ｄ類放大器才被發展出來，相較於類比放大器，他能在低失真情況下進行音訊放大，同時能運作在低功率，逐漸成為音訊放大器的主流；然而，在他放大程序中，仍有一些運算有改進空間。 另一方面，在影像處理方面，由於網際網路的普遍，人們往往運用網路來傳輸影像。而在其傳輸過程中，常會有一些雜訊產生，過去人們使用中值濾波器來消除這些雜訊。但中值濾波器往往會模糊原始影像，同時無法有效消除雜訊。因此需要一個有效率的方法去去除這些雜訊同時能保存原始影像。有不少技術被提了出來改善中值濾波器。 在這篇論文中，將會分成兩個部份，在第一個部份，我們將針對Ｄ類放大器完整程序來進行研究。在Ｄ類放大器程序中，首先是必須產生PWM訊號，因此PCM訊號與PWM訊號之間的轉換是必要的。傳統上我們會使用線性內插演算法來估測，但線性內插演算法會產生一些失真，為了改善這個問題，有些演算法被提出來，我們將會對這些演算法提出些簡短介紹，這些演算法提供我們論文一些概念，我們將發展出一個新的PCM-PWM轉換演算法。它基本原理是利用PWM的增頻現象。利用拋物線來簡單完成增頻，以大幅改善放大器的失真。實驗證明，利用我們演算法在Ｄ類放大器，可有效改善失真率，同時保留線性內插演算法的低功率消耗。 另外，我們將針對影像還原技術做研究，在本篇論文第二部分，我們首先將在次分析脈衝雜訊特性與中值濾波器原理，並學習一些影像還原影算法，最後，我們將提出一個新的雜訊濾除技術並進一步推廣到能處理彩色影像。我們針對影像偵測提出兩個新演算法，他們分別叫做小間偵測演算法和最小邊界中值偵測演算法；其中小間偵測演算法可以偵測出高峰值脈衝型雜訊，即使在高雜訊狀態，而最小邊界中值偵測演算法可以偵測出任何種類的脈衝型雜訊。他們我用的原理都是建構在脈衝型雜訊的特徵。另外，我們提出一個新的技術，動態中值還原演算法，去消除之前被偵測器偵測到的雜訊。他能有效地消除雜訊，即使在雜訊很高的情況下。

As the multimedia getting more and more common and popular, many color images and video are largely produced and transmitted everyday and everywhere. For audio part, because original audio signal is so small and must be amplified. However, there is usually distortion at traditional analogy amplifier. Until 1970’s, new amplifier technique, which is Class D amplifier, has been developed. Class D audio amplifier has gained preponderance over analogy amplifier. Never the less, for some operations, there is still the room for improving. On the other hand, for image processing, people are usually transfer video by using net due to popularity of the Internet. At the process of transmission, there is often some noise. In the past, people is use median filter for removal noise. But median filter may blur the image and it can’t effectively eliminate noise for heavy corrupted image. It is necessary to propose an effective method for removal noise while retain image details. Some techniques are proposed for improving median filter. We divide this thesis into two parts. In part one, we first introduce the complete Class D amplifier processing. The first step of Class D amplifier processing is PWM process. The PCM-PWM conversion is necessary, but we can’t directly covert the PCM signal to NPWM signal. In tradition, we are use linear interpolation to estimate NPWM signal and there is exist distortion problem. For improve the performance of PCM-PWM conversion, some conversion algorithm is proposed. We will introduce those previous methods for conversion. These methods are very helpful to our thesis investigation. We will propose our algorithms for PCM-PWM conversion, including two algorithms. Our algorithm is based on up- sampling. Experimental result shows Class D amplifier by using PWM, which produced by our proposed parabolic interpolation method drastically, reduces crossover distortion while maintaining most of the linear interpolation efficiency. In addition, we will analyze for image restoration. In the part two, we will review the characteristic of impulse noise and the principle of median filter. To study some Detect Noise Algorithm, and in last, we will propose a new De-Noising technique and further generalize this technique for color image processing. For Noise Detection, we propose two new methods, which are named as Gap Detection and Minimum Edge Median Detection. The former detects strong impulse noise in highly corrupted images and the latter can detect any type noise. Out detection is based on gap values, which are obtained through exploiting characteristics of impulsive noise. In addition, we present a new techniques, Dynamic Median Restoration, to remove noise that is detected by our noise detection. It is efficient, very fast, and can remove noise even when noise ratio is very high.