Please use this identifier to cite or link to this item: http://hdl.handle.net/11455/6213
標題: 樹狀向量量化於醫學影像壓縮之應用
Medical Image Compression Using Tree-Structured Vector Quantization
作者: 謝尊伍
Hsieh, Tuzen-Wuu
關鍵字: 醫學影像壓縮;Medical Image Compression;樹狀向量量化;裁減枝葉樹狀向量量化;碎形影像壓縮;影像檔案及通訊系統;Tree-Structure Vector Quantization (TSVQ);Pruned Tree-Structure Vector Quantization (PTSVQ);Fractal Image Compression;Picture Archiving and Communication System (PACS)
出版社: 電機工程學系
摘要: 
隨著科技的進步,數位影像已廣泛地被應用在醫學領域上。一些醫學影像模組已產生純數位的影像資料,例如電腦斷層掃描、核磁共振等等,其餘的類比醫學影像模組亦被數位化。相較於傳統的類比影像,數位影像更較其有利於我們做各種影像處理,比如數位化格式的醫學影像可保存較長久的時間及影像網路傳輸等;也有大量的輔助工具可以增加診斷的可靠度,如影像增強、觀看3-D組織、手術前計劃等等。雖然近年來數位資料儲存設備與通訊頻道容量的能力均有大幅提高,但由於大量的醫學影像資料使得一些演算法的設計變得較為複雜而產生應用上的瓶頸。通常一張影像的大小可達4k*3k*12或18MB,而這些影像的大量資料會使的一些遠端醫療等應用的品質降低,因此醫學影像壓縮仍是一重要課題。
無失真壓縮可經由反轉回復的編碼實現,如Huffman coding, Arithmetic coding, Lempel-Ziv等等。對於醫學影像而言,無失真壓縮法因能夠避免損及診斷的風險而深受喜愛,然而其壓縮率相當有限 (大約4:1或更少),故無法應用在許多範圍。在本論文中,我們研究探討一些以樹狀向量量化概念為基礎的影像失真壓縮方式,藉此應用在醫學影像上。希望影像在不損及我們所需求的資訊主體的失真程度,能得到較高壓縮倍率的編碼方法。所有的編碼方法以一些乳房 光片為實驗對象經由電腦模擬測試。首先,我們經預測器估測影像,然後將預測誤差轉換編碼。為降低編碼計算的複雜度,我們對這些預測誤差採用一種變化比率之裁減枝葉樹狀結構向量量化方式建立編碼簿。我們將檢驗兩種不同的預測方案;第一種方案是以假設高斯分布之線性預測器,在實驗中包含開迴路和閉迴路兩種預測方式。第二種方案是將碎形影像編碼視為預估器。這些壓縮方法研究可以運用在影像檔案及通訊系統(PACS, Picture Archiving and Communication System)中的醫學影像壓縮。

Digital images are used widely in medical practice as the technology advances. Several imaging modalities already produce genuinely digital output, for example computerized tomography (CT) and magnetic resonance (MR); others produce analog images that are subsequently digitalized. Digital images have several advantages over the conventional images. The digital format allows of storage over a significantly longer period of time and transmission of images through network. With digital images, lots of tools can be designed to increase the diagnostic power, such as image enhancement, 3-D view of the body structure, and preplanning tools. Although the performances of the storage devices and the capacity of communication channel have increased a lot recently, the size of the medical image can sometimes increase the complexity of the algorithms. The medical image may have the size of 4k3k12 bits or 18 M bytes per image. It will degrade the performance of certain applications for remote diagnosis. Therefore medical image compression is still an important topic.
Lossless compression can be achieved by an invertible code such as Huffman, Lempel-Ziv, or Arithmetic code. For medical images, lossless compression is preferred since it can avoid the risk of impairing diagnoses. The compression ratios it provides, however, are quite limited (typically about 4:1 or less) and are inadequate for many applications. In this paper, based on the concept of tree-structure vector quantization (TSVQ), we study several lossy compression methods for medical images. Our goal is to obtain high compression ratio without impairing the region of interest (ROI) of images. All the encoding schemes were tested via computer simulations on a set of mammograms. We first estimate the image by a predictor and then encode the prediction error. In order to reduce the computational complexity, we adopt a variable-rate method, Pruned Tree-Structured VQ (PTSVQ), to design codebook for the prediction error. Two different prediction schemes are examined. The first one is a linear predictor with the assumption of Gaussian distribution. Both open-loop and close-loop approaches are included in the experiments. In the second scheme, fractal coding is used as a predictor. These approaches can be used as the compression methods for medical images in the picture archiving and communication system (PACS).
URI: http://hdl.handle.net/11455/6213
Appears in Collections:電機工程學系所

Show full item record
 

Google ScholarTM

Check


Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.