本論文提出一個容忍JPEG失真壓縮之資訊隱藏方法。我們的方法除了能隱藏大量的資訊於掩護影像（Cover image）內，產生的偽裝影像（Stego-image）與掩護影像於外觀上並無太大差異，從視覺觀察無法分辨兩者之間的差別，攻擊者亦無法偵測出該偽裝影像隱藏機密資料。由於數位影像需要佔用大量的儲存媒體，一般傳輸或儲存上皆會採用例如JPEG壓縮法等失真壓縮技術以減小影像之儲存空間，我們的方法產生的偽裝影像，可以使用JPEG來進行失真壓縮處理，增加實用上的價值，但即便如此，從壓縮過並解壓縮後的偽裝影像中萃取出的機密資料，仍然可以達到百分之百的正確率。這對如文、數字與程式等萃取出來有任何一位元發生錯誤將造成意義失真或不可執行的隱藏資料而言是非常重要的。本論文所提出之方法，提供這類資料一個安全隱密的傳輸管道，而這是過去學者們的方法所較少考慮的特質。 In this paper, we present a new steganographic scheme used to embed secret text data into color images. Not only preserve the visual quality, we can use JPEG to com-press the stego-image. JPEG is the most common used image compression technique which reduces the image size drastically without losing the image quality seriously. There are many steganographic scheme have been proposed, but most of them can’t tolerate JPEG compression. That is, after JPEG compressing, the hidden data will be damaged critically. If the embedding data are pictures, accuracy losing would be ac-cepted. But for text data embedding approaches, even one bit error will lead the ex-tracted data to be totally useless. In our scheme, the hidden data are encoded by error correction technique first. Then we embed the encoded data in the spatial domain di-rectly by adjusting the pixel relation in each 3×3 image block. We also adopt the concept of error diffusion method used in halftone technique to improve the image quality to human visual system. By this way, the extracted data from a compressed stego-image could be completely correct.
