Please use this identifier to cite or link to this item: http://hdl.handle.net/11455/5984
標題: 基於相鄰關係之小波無失真影像壓縮
Lossless Wavelet Image Compression Based on Context Relationship
作者: 李舜智
關鍵字: EZW;EZW;SPIHT;wavelet transform;context-based model;lossless;multiresolution;SPIHT;小波轉換;基於相鄰關係模組;無失真;多解析表示
出版社: 電機工程學系
摘要: 
近幾年來已經顯現出小波基底影像壓縮方法對於漸進式的傳輸性能、壓縮效率以及頻寬利用,皆比傳統的方法提供更多的優點,而Embedded Zerotree Wavelet(EZW)編碼技巧以及其推展而出的編碼技術-Set Partitioning In Hierarchical Trees(SPIHT),對於小波基底壓縮方案有著更優良的性能,而且EZW架構示範了更佳的壓縮比例和更簡化的計算複雜度。
基於EZW影像編碼架構包含了以下的三個部分,
1. 可逆的小波離散轉換。
2. 基於相鄰關係模組。
3. 分等級的選擇小波係數使用EZW資料架構。
這種結構提供了幾種可以決定的選擇:
1.在第一階層,我們可以選擇自己所想要的無失真小波濾波器。
2.在第二階層,我們可以選擇一個基於相鄰關係的模組,以利繼續編碼。
3.在第三階層,我們可以選擇一個適當的方法去排列選擇小波係數。
然而在第一階層時,小波轉換把影像的相關度減低,並且提供了一個影像多解析表示。並且在第二階層時,我們會找出小波係數間的關係,並且依其關係找出適當的模組,在最後一階層時,我們會對於小波係數對於影像的重要性去選擇編碼,來繼續編碼。
並且由實驗結果可以發現,對於擁有大部分低頻成分的圖,經由我們所提出來的交叉相鄰關係或稱為z相鄰關係模組以及先做上下與左右比較的相鄰關係模組,則可以得到較佳的壓縮結果。而對於高頻部分佔較多的圖,則是經過先做上下與左右比較的相鄰關係模組,有者較好的壓縮效果。

Wavelet-based image compression techniques offer several advantages over traditional techniques in terms of progressive transmission capability, compression efficiency, and bandwidth utilization. The coding technique SPIHT(Set Partitioning In Hierarchical Trees), which was developed from EZW(Embedded Zerotree Wavelet), has good performance and reduced complexities among wavelet-based schemes.
The framework for EZW-based image coding consists of three components:
1. reversible discrete wavelet transforms,
2. context-based models,
3. ordering and selection of wavelet coefficients using the EZW data structure.
This framework presents several choices to be made:
1. In the first stage, we can choose the lossless wavelet filters.
2. In the second stage, we can use the context-based models.
3. In the third stage, we can choose a suitable method to order and select wavelet coefficients.
In the first stage, the wavelet transforms decorrelate the image data and offer a multiresolution representation of the image. In the second stage, we will find the correlations between wavelet coefficients to look for the best model. And in the last stage, we will encode the data according to the importance of wavelet coefficients. From the experimental results, for images which have mostly low frequency components, we find that the following two models get better compression ratio: the model which adds and substracts the neighboring pixels and the method which selects context model based on the neighboring pixels. For the images which have more high frequency components, the methods which select context model based on the neighboring pixels have better compression efficiency.
URI: http://hdl.handle.net/11455/5984
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.