漸進式影像分享

Abstract

在本篇論文中，我們將機密分享的概念運用在數位影像上。將影像依照低、中、高三個頻率，採取不同的門檻值進行分享；當收集愈多份分存，所得的資訊也會隨之增加；因此，所還原的機密影像品質將會愈佳。我們同時利用影像的特性，有效地將分存的尺寸減小；即分存的尺寸會比原先的機密影像小。 由於產生的分存在視覺上像是雜訊，為了避免駭客的注意，我們採取資料隱藏的技術，將分存藏於載圖之中。所提出的資料隱藏方法，保證像素的誤差在8以下；並使用移位的方法，進一步提升影像PSNR值。 由於資料在網路上傳輸，可能會因為竄改或雜訊，而使得分存受到破壞。因此，我們提出一系列的檢驗與更正的機制，來檢查並處理竄改或雜訊帶來的破壞。我們所提出的檢驗與更正機制，會隨著收集到的分存數增加，檢驗與更正的能力會隨之增加。同時，我們也提出一套機率式的輔助方法，當有愈多正確的分存加入，其更正的準確性也會提升。

In the thesis, we apply the idea of secret sharing to digital images in a progressive way. We subdivide the secret image into low, middle, and high frequency bands; then use different thresholds to share the frequency bands. Therefore, when we collect more shadows, we can get more information; hence, the quality of the reconstructed secret image will be increasing. We also utilize the property of images to reduce the size of shadow, i.e. the size of shadow will be smaller than the size of secret image. Because the shadow is noise-like, to avoid the hacker’s attention, we use the technique of data hiding to embed the shadows in stego images. The proposed data hiding algorithm guarantees that the distortion between the cover image and stego image is at most eight. We also introduce three displacement methods, each of them can increase the PSNR of the stego images. Since shadows might be damaged during transmission because of fakery or noise, we propose a system of error detection and error correction to verify and identify the damage caused by fakery or noise. The ability of detection and correction of the proposed method will increase with the number of collected shadows. A probability –kind error correction procedure is also proposed to handle the case when we do not get enough support shadows. Its accuracy increases as the number of support shadows grows.