CAJ | 학술논문

阐述了一种利用图像内在统计特性进行图像拼接检测的方法.图像拼接检测是一个基于拼接图像特征的模式识别问题.本文通过对游程长度和图像边缘的统计特性进行分析,说明了拼接操作所引起的图像像素不相关性和不连续性.为进行拼接检测,从图像游程长度和图像边缘统计特性中提取特征量,以此特征量去训练神经网络作为最后的分类器.结果表明,由此特征量作为图像拼接检测的标准,检测结果的精确度良好.
천술료일충이용도상내재통계특성진행도상병접검측적방법.도상병접검측시일개기우병접도상특정적모식식별문제.본문통과대유정장도화도상변연적통계특성진행분석,설명료병접조작소인기적도상상소불상관성화불련속성.위진행병접검측,종도상유정장도화도상변연통계특성중제취특정량,이차특정량거훈련신경망락작위최후적분류기.결과표명,유차특정량작위도상병접검측적표준,검측결과적정학도량호.
This paper described a blind image splicing detection method based on the inherent statistical features of images. Image splicing detection is a pattern recognition problem based on the features of splicing images. In this paper, by analyzing the statistical features of run-length and the image edge, indicated the discontinuity caused by image splicing. For splicing detection, characteristics was extracted from the statistical features of image run-length and image edges, then used the characteristics to training the neural network and took the neural network as the last classifier. The final results showed that the characteristics are good for splicing detection.