Please use this identifier to cite or link to this item:
Fractal approach in image coding and texture analysis: the improvement of searching and its applications
|Authors: ||Hsiao-Wen Tin|
|Contributors: ||NTOU:Department of Electrical Engineering|
Fractal;Fractal dimension;Fractal image coding;Image denoising;Network traffic;Side-scan sonar image
|Issue Date: ||2013-10-07T03:01:46Z
|Abstract: ||古典幾何長久以來被用來描述描述自然界物體形狀及結構，然而，大自然的混沌(chaos)現象隨處可見，這些複雜、不規則、且不可微分的形體，很難以古典幾何中的方式精確描述，若無法使用合理的方法描述這些物體，則無法成為自然科學研究的對象。碎形最大的貢獻之一是能對大自然中大部分不規則、碎裂及複雜型態之特徵給予適當之量化描述，使得這些形體得以被研究。 自我相似性是碎形中的重要屬性之一，自我相似性的概念在數學和物理中扮演很重要的角色。經由“迭代函數系統（IFS），可以近似自然物件和圖像。這個概念及方法也建立了碎形影像編碼的基礎。近年來，碎形影像編碼也成了一個廣泛討論的研究主題。碎形編碼通常被認為是一種有效的圖像壓縮方法，主要是由於，碎形編碼是一種具失真性質的編碼方法。碎形編碼最主要的缺點是，在搜尋處理中需要耗費許多時間，因此，改善碎形編碼中的搜尋處理效能是本研究的主要的目標之一。為了克服搜尋速度的瓶頸，我們精心設計了三個減少搜尋區塊的數量的方法。在我們的研究中將透過所提出的方法，改善碎形影像編碼中搜尋過程的效能，以達到減少碎形編碼資料，並維持影像的解碼後品質。 如何去盡可能的消除或減少影像雜訊，並同時維持圖像的紋理信息，是降低雜訊處理中一個困難的問題。本研究提出了一種基於碎形小波編碼的方法，在碎形編碼的搜尋過程中使用測量圖像的粗糙度的方法做為搜查的比較基準，由實驗結果可知，我們所提出的方法是很適合用於側掃聲納圖像降低雜訊，並且可以維持影像的可視品質。 碎形維數（FD）是碎形性質的抽象度量，並可以做為提取特徵的指標。在我們的研究中，基於碎形的尺度不變的性質，從看似複雜的網路流量的變動，發現其變化的有序性，我們提出一個應用碎形維度測量網路流量的方法。|
Euclidean geometry describes the artificial shapes well since its development. Natural science also applied the Euclidean system to describe the shapes and structures of natural objects for a long time. However, chaos exhibits in nature everywhere. Chaos objects are characterized as complex, irregular and non-calculable which is hardly described by Euclidean system. An indescribable object is not able to be a research object. Chaos objects remained as indescribable until fractal geometry was presented. The main contribution of fractal geometry is that it well describes quantitatively most of the irregular, broken and complex natural objects. Self-similarity is one of the important features of fractal in both mathematics and physics. In practical, iterated function system (IFS) was adopted by the fractal coding to approximate the natural objects and images. In recent years, IFS method has become widely popular in fractal image coding research. Such fractal image coding is usually considered as an efficient image coding method primarily due to it is a lossy compression in nature. However, it requires some heavy computations due to the fractal block encoding requires much of the processing time to search the domain blocks, which is the major hurdle for performance. To overcome the performance bottleneck, this dissertation proposed three fractal coding algorithms based on reducing the number of domain blocks in the domain pool. The elaborately designed approaches improve the searching efficiency and preserve the image quality after encode. Image denoising is a difficult problem which is how to detract image texture information as little as possible while removing or reducing noise as much as possible. Fractal-wavelet coding methods were an efficient image denoising method in various areas but not in texture analysis. This dissertation proposes an elaborative approach to measure the image roughness for texture analysis and then introduces to Fractal-wavelet (FW) coding process. The experimental results indicate that the proposed algorithm is adaptable in denoising side-scan sonar images and that the images are more appealing visually. The fractal dimension (FD) is an abstract measure of fractal property and used to index the extracted feature. From analyzing the network traffics in various scales, this study observed the very complicated variances of network flows were actually fractal. Therefore, this dissertation proposes an approach to quantify the bursty traffic that degrades network performance and reliability by using fractal dimensions.
|Appears in Collections:||[電機工程學系] 博碩士論文|
Files in This Item:
All items in NTOUR are protected by copyright, with all rights reserved.