臺灣博碩士論文加值系統

在許多現今的影像壓縮編碼系統中,離散餘弦轉換 (Discrete Cosine
Transform) 技術被廣泛的使用,而結合人類視覺系統 (Human Visual
System) 與離散餘弦轉換的編碼系統,除了具有離散餘弦轉換編碼的良好
編碼效能之外,更能提升重建影像在視覺上的品質。在本篇論文中,我們提
出一個以人類視覺系統為基礎的可調變式離散餘弦轉換編碼系統。此系統
主要由兩個部分所組成,包括量化器位階調整因子函數 (Range Scale
Factor Function) 的產生及去除區塊效應。前者產生一個量化器位階調
整因子,此因子能根據不同區塊中不同的視覺上的特性,而給予不同程度的
量化,以達到產生高品質重建影像的目地。後者可以消除在轉換過程中因
區塊失真 (Distortion) 程度不一而產生的區塊效應。實驗結果顯式此系
統在 0.4 bit/pixel 人滿意的重建影像。

An adaptive cosine transform coding scheme for gray-scale
images based on HVS is proposed in this thesis. Ngan, Leong,
and Singh's approach is adopted in this paper. In our block
adaptive scheme, a new range scale factor function is proposed
to improve the subjective quality of those blocks which contain
edges. Besides, we will propose a new adaptive block-effect
redunction method to reduce the block effect resulting from
block transform coding schemes. This scheme is combined with
perceptual distortion minimization and two-dimensional block
distortion equalization algorithms. Experimental results show
that the subjective quality of the reconstructed images at a
bit rate of 0.4 bit/pixel is very appealing.