根植於動態向量量化像素差值修正之新影像壓縮技術

A New Image Compression Technique Based on Pixel Differential Correction of Dynamic Vector Quantization

吳家禎(Chia-Jen Wu)

指導教授 ： 蔡正發

國立屏東科技大學/管理學院/資訊管理系所/碩士(2016年)

摘要

本文提出基於空間域LBG的改良式影像壓縮技術。改良成效將以實作影像壓縮結果，與頻率域的JPEG作一系列「品質—壓縮比」的比較。
此技術核心的演算概念為：
1. 像素差值修正：在完成傳統LBG壓縮編碼後，利用原始影像與壓縮解碼後影像之間存在的像素差值，進行差值修正。每修正一次，兩者的誤差值就會減少，PSNR值因而提高，但同時也因為增加額外的編碼簿及索引表，造成壓縮比下降。
2. 動態向量量化編碼：將修正過程的編碼簿大小，設定為原編碼簿大小的四分之一。若原編碼簿使用256的大小，則修正編碼簿僅使用64的大小，以減少壓縮效率的損失。
此外，影像前置處理部分，使用不重要位元刪除及還原技術，以增加壓縮比。後置處理則使用差分編碼搭配霍夫曼編碼的位元儲存壓縮技術。最後將上述所有演算法整合起來，進行影像壓縮實作。

關鍵字

像素差值修正 ； 動態向量量化 ； 不重要位元刪除 ； LBG演算法 ； 差分編碼 ； 霍夫曼編碼

並列摘要

In this thesis, we propose new image compression techniques based on LBG in spatial domain. The real effects of these techniques will be compared by a series of image quality – compression ratio evaluations with JPEG which is compressed in frequency domain.
The algorithm concepts of the core techniques are:
1. Pixels differential correction：After finishing traditional LBG encoding, we amend the deviations between original images and decoding images. As the amending process is ongoing, the differential values are getting lower and PSNR value is getting higer. But the extra amending codebooks and index tables will occupy storage space at the same time.
2. Dynamic vector quantization encoding : the amending codebook size will be set as a quarter of original codebook size in order to improve compression ratio. For example, if we use 256 codebook size for encoding, the amending codebook size will be 64.
Furthermore, we use least significant bit deletion and reduction techniques in the image preprocess and delta plus Huffman coding methods will be applied in the post process in order to increase compression efficiency.
Finally, all the algorithms illustrated above will be integrated to implement image compression.

並列關鍵字

Pixels Differential Correction ； Dynamic Vector Quantization ； Least Significant Bit Deletion ； LBG Algorithm ； Delta Coding ； Huffman Coding

參考文獻

[4] 林于峻，具新的高效能與高效率之增強型自組織特徵映射圖於影像壓縮問題之研發，國立屏東科技大學資訊管理學系碩士論文，2009。

[10] Braquelaire, J. P. and Brun, L., “Comparison and optimization of methods of color image quantization,” IEEE Transactions on Image Processing, Vol. 16, No. 7, pp. 1048-1052, Jul. 1997.

[11] C. C. Chang, D. C. Lin, and T. S. Chen., “An Improved VQ Codebook Search Algorithm Using Principal Component Analysis,” Journal of Visual Communication and Image Representation, Vol. 8, No. 1, pp. 27-37, 1997.

[12] C. C. Han, Y. N. Chen, C. C. Lo, and C. T. Wang.,” A Novel Approach for Vector Quantization Using a Neural Network, Mean Shift, and Principal Component Analysis-based Seed Re-initialization,” Signal Processing, Vol. 87, pp. 799-810, 2007.

[13] C. M. Huang and R. W. Harris., “A Comparison of Several Vector Quantization Codebook Generation Approaches,” IEEE Trans. on Image Processing, Vol. 2, No. 1, pp. 108-112, 1993.