CAJ | 학술논문

提出一种使用Harris 兴趣点的拷贝检测方案.首先通过Harris检测器提取图像的局部特征点,然后构建以Harris兴趣点为中心的多重圆环区域,并计算圆环区域的信息熵,作为特征向量,通过比较特征向量之间的距离,判断局部领域的相似性.最后通过特征点的全局一致性检测,判断检测图像是否为原始作品的拷贝.实验表明该拷贝检测方法对图像的几何变换具有很强的鲁棒性,同时也能抵抗大多数的信号处理攻击.
제출일충사용Harris 흥취점적고패검측방안.수선통과Harris검측기제취도상적국부특정점,연후구건이Harris흥취점위중심적다중원배구역,병계산원배구역적신식적,작위특정향량,통과비교특정향량지간적거리,판단국부영역적상사성.최후통과특정점적전국일치성검측,판단검측도상시부위원시작품적고패.실험표명해고패검측방법대도상적궤하변환구유흔강적로봉성,동시야능저항대다수적신호처리공격.
An image copy detection scheme is proposed by using Harris interest points.Firstly,the Harris interest points are extracted and several small regions around the interest point are located as image patches.Then each patch is partitioned into several concentric cirques,and the entropy information vectors are extracted from each image patch.And then,the similarity measurement is done by comparing entropy information vectors of two regions.Finally spatial distribution is detected by using the location of interest points to decide if one image is a copy or not.Integrating interest points and entropy information makes the method be robust to both geometric attacks and signal processing.