Scholarlink

<p>Image forgery detection has become a critical challenge in digital forensics, copy-move forgery being one of the most widespread manipulation techniques. Copy-move fabrication is a widespread method of image manipulation in which a specific segment of an image is replicated and superimposed onto another region of the image. Traditional keypoint-based methods using SIFT and FLANN matching suffer from false correspondences and unbalanced clustering thresholds, restricting the precise localization of forgeries. We proposed a novel Adaptive Density-Spatial Clustering (ADSC) optimization integrated into a SIFT, FLANN, RANSAC framework. The proposed method dynamically adjusts clustering parameters using local displacement vector density, overcoming fixed-threshold limitations in conventional DBSCAN-based clustering. Experimental validation on the MICC-F220, MICC-F2000, and IEEE image forgery datasets demonstrates consistent improvements in pixel-level Intersection over Union (IoU), precision, and recall. The proposed framework provides improved robustness, accurate localization. The system proves highly robust against geometric transformations, including rotation and illumination variations, while maintaining computational efficiency for varying forged region sizes.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Efficient detection and localization of single and multiple copy-move forgeries in digital images

  • 期刊论文

  • 发表日期：

    2026年7月27日

查看全文
  • Urmila Samariya,
  • Shailesh D. Kamble,
  • Subhash Singh

摘要

Image forgery detection has become a critical challenge in digital forensics, copy-move forgery being one of the most widespread manipulation techniques. Copy-move fabrication is a widespread method of image manipulation in which a specific segment of an image is replicated and superimposed onto another region of the image. Traditional keypoint-based methods using SIFT and FLANN matching suffer from false correspondences and unbalanced clustering thresholds, restricting the precise localization of forgeries. We proposed a novel Adaptive Density-Spatial Clustering (ADSC) optimization integrated into a SIFT, FLANN, RANSAC framework. The proposed method dynamically adjusts clustering parameters using local displacement vector density, overcoming fixed-threshold limitations in conventional DBSCAN-based clustering. Experimental validation on the MICC-F220, MICC-F2000, and IEEE image forgery datasets demonstrates consistent improvements in pixel-level Intersection over Union (IoU), precision, and recall. The proposed framework provides improved robustness, accurate localization. The system proves highly robust against geometric transformations, including rotation and illumination variations, while maintaining computational efficiency for varying forged region sizes.