English  |  正體中文  |  简体中文  |  Items with full text/Total items : 43312/67235
Visitors : 2107302      Online Users : 18
RC Version 5.0 © Powered By DSPACE, MIT. Enhanced by NTU/NCHU Library IR team.

Please use this identifier to cite or link to this item: http://nchuir.lib.nchu.edu.tw/handle/309270000/154565

標題: 極低失真與無失真之高動態範圍影像資訊嵌入演算法
A Study of Low Distortion or Distortion-free Data Embedding Algorithms for High Dynamic Range Images
作者: 柯衍慶
Ke, Yen-Ching
Contributors: 王宗銘
Chung-Ming Wang
資訊網路多媒體研究所
關鍵字: 高動態範圍;無失真資訊嵌入;極低失真資訊嵌入;影像註記;藏密學;多群機制
high dynamic range images;distortion-free;very low image distortion;triplet coding;pixel cluster mechanism;image annotation;image steganography
日期: 2012
Issue Date: 2013-11-21 10:56:54 (UTC+8)
Publisher: 資訊網路多媒體研究所
摘要: 隨著資訊與網路的發達,資訊嵌入這項技術在資訊安全的議題中有越來越被重視之趨勢,由於高動態範圍影像所能夠表現出的豐富色彩遠遠的大於低動態範圍影像,所以高動態範圍影像有日漸成為主流影像的趨勢。因此本論文針對高動態範圍影像提出極低失真與無失真的資訊嵌入演算法。
我們所提的第一個演算法為「高動態範圍影像三區間編碼資訊嵌入演算法」,簡稱TRICOD。我們充分善用所有高動態範圍影像的同質性像素提供之狀態數,加上三區間編碼的技巧來進行資訊的嵌入,故此演算法可以在不提高影像變動量的前提下,增加訊息嵌入量。實驗結果顯示:我們所提出的演算法相較於最新文獻,可提升5.52%~5.79%之資訊嵌入量。高動態範圍偽裝影像經色調映射後產生的低動態範圍影像,經量化與視覺化差異量測(HDR-VDP)並無影像失真。
我們所提的第二個演算法為「使用Null Pixel的高動態範圍影像的資訊嵌入演算法」,簡稱NULPIX。我們發現先前學者所提出的高動態影像像素分類中,尚有一類像素可以使用,只要針對分類為Null之像素的E通道來進行,即能夠完成嵌入資訊演算法。實驗結果顯示:相較於最新的文獻,NULPIX演算法平均可以提高48.37%之嵌入量,高動態範圍偽裝影像經由色調映射後之影像,經量化或視覺化差異量測下並無影像失真。
我們所提的第三個演算法為「高動態範圍影像最低失真資訊嵌入演算法」,簡稱FEAPIX。此演算法是文獻首創的演算法。我們提出一個最佳化計算機制,將原本無法被使用的像素,經由最佳化的計算後,使得資訊能夠順利的嵌入。實驗結果顯示:演算法可大幅提高2.77~3.02倍之秘密訊息嵌入量。高動態範圍影像色調映射後之影像經由量化後其PSNR仍高達76.60~84.44 dB;影像雖略有些微失真,但經由人眼視覺化差異量測的演算法得知,仍無法辨識差異性。
我們所提的第四個演算法為「多像素群組機制的高動態範圍影像資訊嵌入演算法」,簡稱CLUTCOD。我們利用同質性像素排列組合之特性並配合三區間編碼的技巧,使用M個像素(M>=2)當作一個群組來嵌入訊息,故可以在不增加偽裝影像變動量的前提下,提升資訊的嵌入量。實驗結果顯示:演算法可提升317~2397個位元的資訊嵌入量。我們使用色調映射技術將高動態掩護與偽裝影像轉換成低動態掩護與偽裝影像,以供量化差異與視覺化差異分析。量化分析結果顯示:低動態掩護影像與偽裝影像間並無PSNR數值的差異;我們使用視覺化差異預測器(HDR-VDP)來量測低動態掩護與偽裝影像之視覺化差異。視覺化差異分析所產出之機率地圖顯示:掩護與偽裝影像之間,被人眼察覺有差異之機率為零。
本論文主要的貢獻有下列四項:第一、使用三區間編碼的技巧善加利用所有的狀態數,改進先前無失真演算法的侷限嵌入量,進一步提高資訊嵌入藏量。第二、針對之前從未被使用過的像素,提出了一種全新的演算法,使其能夠被善加利用,提高嵌入量。第三、提出最佳化計算機制,充分利用原先無法被使用的像素,能順利的嵌入資訊,雖造成影像有極低的失真,但仍可進一步提高資訊的嵌入量。第四、對於三區間編碼的技巧,我們以多像素為一群來嵌入訊息,在不造成失真情況下,進一步提升嵌入量。本論文提出的四個演算法擴大高動態範圍影像的影像註記與藏密學應用之範疇。
In this paper, we investigate data embedding algorithms for high dynamic range images encoded by the RGBE image format. We present four algorithms that have the distortion-free feature and one algorithm that demonstrates the feature of very low distortion.
Our first algorithm belongs to the distortion-free manner. In this algorithm, we make use of all statuses produced by the pixel variation and employ triplet coding technology to increase the embedding capacity. Comparing with the previous work, our algorithm can improve the embedding capacity in the range between 5.52% and 5.79%. No image distortion is encountered when tone mapping the high dynamic range embedded images to produce the low dynamic range embedded image.
The second algorithm we introduce belongs to the distortion-free manner. In this algorithm, we take advantages of the “null” pixel, a new pixel category produced by the E channel, where we embed messages into these pixels to expand the embedding capacity. Experimental results show that comparing to our counterparts, our algorithm can offer an average of 48.37% embedding capacity without causing any image distortion.
The third algorithm we develop belongs to the very low distortion manner. We adopt an optimization computation mechanism for the R, G, B channels to generate a number of potential pixels, referred to as “promising” and “feasible” pixels. These pixels cause the least image distortion when operating the message embedding. Comparing to our counterparts, the algorithm can largely increase the amount of embedding capacity with the magnitude between 2.77 and 3.02. The tone mapped image presents high PSNR values (76.60~84.44 dB) showing no perceivable visual difference.
The final algorithm we propose is with the distortion-free manner. We take advantage of homogeneous pixel representation and combine a group of M pixels (M>=2) as a pixel cluster to generate sufficient statuses for message embedding. This approach allows us to adopt the triplet coding technique to increase the embedding capacity without incurring any image distortion. We compare our scheme of using 2 pixels as a cluster with previous results of using a single pixel. The comparison indicates that our algorithm can provide larger payloads in the range of 317~2397 bits. We adopt the tone mapping scheme to produce low dynamic range images to quantize the image difference, and we employ the HDR-VDP technique to inspect the visual difference between the cover and stego images. The image difference quantization results show that no distortion is encountered. The probability map produced by HDR-VDP inspection is in grey color indicating that the detection probability of visual difference is null.
In conclusion, our work offers the following four contributions: we exploit the triplet coding technology and increase the capacity for non-distortion algorithm; we make use of a new pixels to convey messages, raising the embedding capacity; we develop the optimization computation mechanism fully using pixels not available in our counterpart to further increase the embedding capacity; we adopt the pixel cluster scheme allowing the increase of concealed messages without causing image distortion. The four algorithms developed are adequate for applications of image annotation and image steganography.
Appears in Collections:[依資料類型分類] 碩博士論文

Files in This Item:

File SizeFormat
index.html0KbHTML138View/Open


 


學術資源

著作權聲明

本網站為收錄中興大學學術著作及學術產出,已積極向著作權人取得全文授權,並盡力防止侵害著作權人之權益。如仍發現本網站之數位內容有侵害著作權人權益情事者,請權利人通知本網站維護人員,將盡速為您處理。

本網站之數位內容為國立中興大學所收錄之機構典藏,無償提供學術研究與公眾教育等公益性使用。

聯絡網站維護人員:wyhuang@nchu.edu.tw,04-22840290 # 412。

DSpace Software Copyright © 2002-2004  MIT &  Hewlett-Packard  /   Enhanced by   NTU/NCHU Library IR team Copyright ©   - Feedback