近年來，由於資訊技術和網際網路的流行，使得數位影像的應用日漸廣泛，也因為數位影像容易傳播、修改，使得少數有心人士容易取得並竄改影像，原始影像的正確性因而受到質疑，造成當事人諸多困擾。有學者提出以離散小波(DWT)的方法將ROI的特徵值嵌入轉換頻率域後的ROB中，在影像應用前自頻率域的ROB位元資料中取出特徵值，並與ROI資料進行比對和復原，以確定影像的正確性，但其缺點在遭竄改部份只限定於ROI區域，ROI大小的設定會受到限制，以及重建影像品質不佳。在本文中我們提出二種不同的改善方法：一、在影像經一階DWT運算後，將低頻區ROI的位元值隨機嵌入三份備份於ROB中頻帶，在影像全區域遭受竄改時，藉由投票方式決定ROI部份的回復。二、將ROI的位元值以JPEG演算法壓縮後，隨機嵌入一份備份於一階DWT後的ROB中頻帶，在影像的ROI區域遭受竄改時，可以在ROB中取出備份來做資料回復。實驗結果顯示我們提出的二種方法在回復影像時，可以在ROI大小限制或重建品質上各有不同程度的改善。

In recent years, as a result of the information technology and the internet popularity caused the applications of digital image to be widespread. The digital image may be obtained and tampered easily. For the accuracy of primary image, there are some scholars proposed the method of the discrete wavelet transform (DWT) to take the features of ROI into the region of transformation-frequency ROB. Take out the features from the frequency region of ROB bit before the image applications, and compare with the ROI data to make sure the accuracy of the image. There are some disadvantages of this scheme. The part of tampering only limited on ROI region, the size of ROI region is small, and the quality of reconstructed image is not well. We propose two different methods to improve the image recovery in this paper. The first scheme takes the low frequency region of ROI bit stochastically insert three backups into ROB of middle frequency in the DWT domain. The part of ROI recovery is determined by the voting from the computer formula when the image region tampered. The second scheme compresses the region of ROI by the JPEG, and inserts the compressed data in ROB of middle frequency stochastically after the DWT. When the image region of ROI tampered, directly recover by backup data of ROB. Simulation results show some improvement of the limit of ROI size or the recovered quality by the proposed methods.