Scholarlink

<p>The identification of image manipulation, specifically copy-move and splicing manipulation, continues to be a significant obstacle in domains like public opinion monitoring, the military, and the media. There is an urgent need for reliable tampering detection systems due to the widespread use of forged images and the damage they cause to public confidence. However, the complexity of forgery patterns and differences in the scale of manipulated regions make it difficult for current techniques to achieve high accuracy. The research paper proposes a robust model for tampered image detection using a DCB-PSO optimizer (deep convolutional network with binary particle swarm optimization) to fill these gaps. Multiple classifiers, such as SVM, KNN, logistic regression, decision trees, and gradient boosting, are integrated with deep convolutional neural networks (CNNs) for feature extraction and binary particle swarm optimization (BPSO) for effective feature selection. Thorough tests on the CASIA v1.0, CASIA v2.0, and NC16 datasets show how effective the suggested approach is; under particular configurations (<i>K</i> = 5, <i>N</i> = 20), KNN achieves the highest accuracy of 97.16% and 98.53% precision. BPSO considerably improves detection performance compared to feature selection-free models. These findings highlight the effectiveness and resilience of DCB-PSO, giving computer forensic specialists a reliable tool to identify image manipulations and stop the spread of false information.</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

2DCB-PSO: An Advanced Image Forensics Technique for Enhancing Accuracy in Digital Image Analysis

  • 期刊论文

  • 发表日期：

    2025年5月22日

查看全文
  • Choudhary Shyam Prakash,
  • Sahani Pooja Jaiprakash,
  • Naween Kumar,
  • Anand Nayyar

摘要

The identification of image manipulation, specifically copy-move and splicing manipulation, continues to be a significant obstacle in domains like public opinion monitoring, the military, and the media. There is an urgent need for reliable tampering detection systems due to the widespread use of forged images and the damage they cause to public confidence. However, the complexity of forgery patterns and differences in the scale of manipulated regions make it difficult for current techniques to achieve high accuracy. The research paper proposes a robust model for tampered image detection using a DCB-PSO optimizer (deep convolutional network with binary particle swarm optimization) to fill these gaps. Multiple classifiers, such as SVM, KNN, logistic regression, decision trees, and gradient boosting, are integrated with deep convolutional neural networks (CNNs) for feature extraction and binary particle swarm optimization (BPSO) for effective feature selection. Thorough tests on the CASIA v1.0, CASIA v2.0, and NC16 datasets show how effective the suggested approach is; under particular configurations (K = 5, N = 20), KNN achieves the highest accuracy of 97.16% and 98.53% precision. BPSO considerably improves detection performance compared to feature selection-free models. These findings highlight the effectiveness and resilience of DCB-PSO, giving computer forensic specialists a reliable tool to identify image manipulations and stop the spread of false information.