Hiding and Inpainting Region of Interest in Images

• Main Authors: Hsien-Jen Hsu, 許獻仁
• Other Authors: Ran-Zan Wang
• Format: Others
• Language: zh-TW
• Online Access: http://ndltd.ncl.edu.tw/handle/19510823908906097406

碩士 === 元智大學 === 資訊工程學系 === 102 === This thesis proposes an image camouflaging method to conceal the region of interest (ROI) of an image. The camouflaged image exhibits content without the ROI to general users, but the authenticated user can detect the embedded information to restore the content of the ROI. In the encoding phase, the ROI is selected manually by the user. A high-payload data hiding method is applied to embed the data of ROI in the background area of image, and the content of ROI is cleared and inpainted to maintain a natural appearance of the whole image. An image verification mechanism is also designed for indicating the existence of embedded ROI in the image. In the decoding phase, authenticated users detect the embedded ROI using the image verification mechanism with correct secret key, and extract the content of ROI to construct the original image. Experimental results show that the generated camouflaged image has natural appearance, and high-quality image with content of ROI can be reconstructed using the proposed scheme. The proposed can be applied in protecting sensitive information such as license plate or tagged private identification information of an image, or utilized to remove the interference objects such as irrelevant visitors or billboards of landscape image. Keywords: image inpainting, image authentication, data hiding and object camouflage.