對於抑制高斯雜訊、脈衝雜訊與壓縮缺陷之研究

Abstract

在現今許多電子消費產品中，高畫質數位影像成為主流的資訊來源，對於影像品質要求也愈來愈高。因此，處理影像中的雜訊便成為首要解決的課題。本論文針對常見的高斯雜訊、脈衝雜訊以及壓縮缺陷，設計幾個簡單的處理方法。首先我們對於未知雜訊類型之影像作壓縮缺陷與脈衝雜訊之偵測，再使用我們所提出的演算法抑制雜訊以提升影像品質。為了符合實際應用需求，對於影像中之壓縮缺陷，我們所設計之處理方法不需任何原始影像壓縮的編碼資訊，僅處理解壓縮後的影像資料，而且演算法不需太多花費計算時間與硬體資源，以適用於彩色影像與視訊的處理。

In many of today's consumer electronics products, high-definition digital images have become a main source of information. As the demand of high quality of images keeps increasing, the reduction of image noise becomes a primary topic. In this thesis, we design several approaches for the reduction of Gaussian noise, impulse noise, and compression artifacts. First, images with unknown type of noise would be processed for the detection of compression artifact and impulse noise. The proposed methods are then used to enhance image quality. For practical applications, the proposed algorithms do not require any coding information, and can be directly applied over decompressed data. The proposed methods require low computational loads and hardware resources, and can be efficiently applied to various kinds of color images and videos.