Abstract:
Based on the research of extracting image texture features using Local Walsh Transform(LWT), this paper proposes an image region copy-move forgery detection algorithm fused with LWT texture features. After dividing the detected image into multiple overlapping blocks with the same size, the algorithm uses LWT to extract an image texture feature vector for each block and estimate the texture complexity of the detected image, which is used to choose the threshold for the determining similar image blocks. It sorts all texture feature vectors in alphabetical order, and calculates the proper threshold value. The algorithm detects and locates the tampered region by predetermined similarity criteria. Experimental results show that the algorithm performs better than the previous detection algorithm based on Principal Component Analysis(PCA) in terms of detection accuracy and false positive rate.
Key words:
forgery detection,
image copy-move forgery,
Local Walsh Transform(LWT),
texture feature,
feature vector,
texture complexity
摘要： 在对局部沃尔什变换(LWT)提取图像纹理特征研究的基础上，提出一种融合LWT纹理特征的区域复制篡改检测算法。将待检测图像分成大小相同的重叠块，利用LWT提取每个图像块的纹理特征，估算整个待检测图像的纹理复杂度，对获得的每个图像块纹理特征向量进行字典排序，并根据估算到的纹理复杂度值选择合适的相似图像块判定阈值，按照预定的相似标准，检测且定位出篡改区域。实验结果表明，该算法在准确率和虚警率方面均优于经典的基于主成分分析法的检测算法。
关键词:
篡改检测,
图像复制篡改,
局部沃尔什变换,
纹理特征,
特征向量,
纹理复杂度
CLC Number: