CAJ | 학술논문

数字图像自嵌入与恢复是在图像中嵌入与自身相关的信息，用于接收端判断图像是否被篡改并恢复被篡改区域的内容。提出一种基于显著性分类的数字图像自嵌入方法，根据图像自身的特点对图像进行分类，动态决定每个区域的参考数据量和嵌入容量，采用喷泉码对参考数据进行编码，并将其嵌入到原图像的不同区域中。与传统方法相比，所提出的方法主要有以下两点优势：在生成参考数据方面，所提出的方法可根据内容分类确定编码长度，在保证整体恢复质量的同时，能重点保护显著性区域；在数据嵌入方面，选择在不同区域中嵌入不等的数据量，可避免传统均匀嵌入法导致图像伪轮廓等缺陷，保证含密图像具有良好的质量。
수자도상자감입여회복시재도상중감입여자신상관적신식，용우접수단판단도상시부피찬개병회복피찬개구역적내용。제출일충기우현저성분류적수자도상자감입방법，근거도상자신적특점대도상진행분류，동태결정매개구역적삼고수거량화감입용량，채용분천마대삼고수거진행편마，병장기감입도원도상적불동구역중。여전통방법상비，소제출적방법주요유이하량점우세：재생성삼고수거방면，소제출적방법가근거내용분류학정편마장도，재보증정체회복질량적동시，능중점보호현저성구역；재수거감입방면，선택재불동구역중감입불등적수거량，가피면전통균균감입법도치도상위륜곽등결함，보증함밀도상구유량호적질량。
The purpose of image self-embedding is to achieve content authentication and self-recovery by imperceptibly embedding relevant information into the host image. This paper proposes a novel self-embedding method based on signiifcance classiifcation. After classifying the image into three kinds of regions dynamically, the code length and embedding capacity of each block are determined. Using fountain coding, the reference information is embedded into the entire image. The proposed method is superior to the traditional methods in two aspects. First, the method assigns different code length to different regions so that the recovery quality for the whole image is good and signiifcant regions can be protected effectively. Second, the method embeds different amounts of data into different regions thus avoiding false contour and ensuring good quality of the stego-image.