International Journal of Computer Applications |
Foundation of Computer Science (FCS), NY, USA |
Volume 186 - Number 62 |
Year of Publication: 2025 |
Authors: Smruti Dilip Dabhole, G.G. Rajput, Rajendra Hegadi, Prashantha |
![]() |
Smruti Dilip Dabhole, G.G. Rajput, Rajendra Hegadi, Prashantha . Image Tamper Detection using the Fusion CMFD Model with Advanced VGG16 Features. International Journal of Computer Applications. 186, 62 ( Jan 2025), 42-51. DOI=10.5120/ijca2025924399
Due to the widespread availability of the internet and the abundance of devices capable of capturing images, there has been a significant increase in the number of images shared online. These images are easily manipulated using advanced software tools like Adobe Photoshop, leading to the creation of fake visuals. As the sophistication of image and video editing tools continues to advance, distinguishing between authentic and altered images has become increasingly challenging. Thus, it is crucial to verify the authenticity of images before deriving any significant insights from them. In this paper we present a novel method for detecting and categorizing tampered and genuine regions within images, without reliance on reference images. The proposed novel approach called ‘Fusion CMFD’ ‘Copy Move Forgery Detection (CMFD)’, model includes fusion of ‘Manipulation CMFD’ model and ‘Similarity CMFD’ model. Features are extracted in both models using VGG16 neural network architecture where, features from convolutional layers are intelligently concatenated to enhance discriminative power and facilitate more accurate identification of genuine and tampered regions within an image. ‘Similarity CMFD’ model employs the VGG16 architecture, utilizing self-correlation to assess feature similarity between the input image and its corresponding mask. Potential features are aggregated using Percentile Pooling, and then a Mask Decoder is utilized to upscale feature maps to the original image dimensions. ‘Manipulation CMFD’ includes feature extraction using VGG16 and mask decoder. The proposed innovative approach promises enhanced accuracy and robustness in detecting tampered and genuine regions within images, opening up new avenues in the field of image forensics and enhancing overall security measures in digital content authentication. Experiments are performed on images from the MICC-F2000 [1] dataset. The results are compared against existing methodologies reported in the literature and cross verified with MICC-F220 dataset. Performance has been analyzed using parameters namely, accuracy, precision and recall.