影像的快速分享與高容量隱藏，及它們在影像修復的應用

Abstract

這是一個三年期計畫。第一年主題是快速影像分享;第二年主題是高容量資訊隱藏 方法之設計與資訊隱藏量上限之探討;第三年主題是快速影像分享法與高容量資訊隱藏 法在影像驗證與修復之應用。 第一年計畫的子題一是使用布林運算的快速機密影像分享法。儲存空間很小，而且 解碼運算量也少。子題二是多項式(t, n)分享方法的快速編解碼。子題三是快速編解碼 的動態權重分享方法。 第二年計畫的子題一是以機密影像之漢明長度為基礎的“無失真”資訊隱藏方法。可 以藏入較高的資訊量，而偽裝影像品質仍然好。子題二是基於機密影像中相鄰像素間相 似性而得的高容量隱藏方法。讓機密影像有一點失真，但機密影像可以是掩護影像的數 倍。子題三探討每張掩護影像的隱藏容量上限。子題四是基於多維像素空間而得的影像 隱藏方法，希望研究出隱藏率接近最小上限的影像隱藏演算法。 第三年計畫的子題一是區塊式影像驗證與自我還原系統。希望利用第二年資訊隱藏 的研究成果，搭配第一年多項式分享法的加速，發展出一個影像驗證與自我還原系統。 子題二為設計出一個針對感興趣區域(ROI)的驗證與自我還原系統。子題三是使用餘數 分享法做影像快速修復。

This is a 3-year project. Year 1 is for the fast sharing of an image. Year 2 is for the design of high-capacity hiding methods, including the study of the upper bound of each cover image’s hiding capacity. Year 3 is for images’ authentication and recovery, by applying the fast sharing methods and high-capacity hiding methods designed in Years 1 and 2. Year 1 has three subtopics, as follows. 1a): the design of a fast sharing method based on Boolean operations. The storage space will be small, and the decoding time will be very short. 1b): to accelerate the polynomial-based (t, n) threshold sharing method. In general, polynomial-based image-sharing method is good in getting small-size shares; however, it is not fast. Therefore, we will analyze its repeated patterns and design a new algorithm to accelerate speed. 1c): fast sharing for a system with shares of dynamic weights. In many companies, people of different importance levels might have different weights when they vote for the disclosure of a shared secret image. The weights might also be dynamic to match the company’s dynamic developing. We will design a fast method for such system. Year 2 is has four subtopics. 2a): a lossless high-capacity image-hiding method based on secret image’s Hamming norm. 2b): a lossy high-capacity image hiding method based on the similarity between neighboring pixels of the secret image. The secret image can be several times larger than the cover image, with small distortion in the recovery of secret image. 2c): an upper bound of a cover image’s hiding capacity. It is a natural barrier for all hiding methods. We will require this upper bound be as low as possible. 2d): a high-capacity image hiding method based on a high-dimensional space for pixels. We will design a new method whose hiding-rate is close to the upper bound found in 2c) above. Year 3 is an application of the sharing methods and hiding methods designed in Years 1 and 2. The three subtopics are 3a): a block-based image authentication method with self-recovery of tampering. The method utilizes the accelerated-polynomial-based sharing of Year 1 and a hiding method of Year 2. The method detects which parts of the image are tampered, and then do automatic repairing. 3b): authentication and self-recovery of an image’s ROI (Region of interest). The image’s ROI will be shared and hidden in the non-ROI of the same image. Later, when the damaged area of the processed image is so large that it covers both ROI and non-ROI, the recovery of the ROI and non-ROI will be processed separately. 3c): using fast sharing in image’s fast-repairing. The given image will be shared by a module-based sharing method so that each share is small and easy to be hidden. After our processing, image will still look natural. Later, if being tampered, the image can still be recovered in a fast speed.