Dual adaptive deep convolutional neural network for video forgery detection in 3D lighting environment
- PDF / 12,363,871 Bytes
- 22 Pages / 595.276 x 790.866 pts Page_size
- 75 Downloads / 253 Views
ORIGINAL ARTICLE
Dual adaptive deep convolutional neural network for video forgery detection in 3D lighting environment V. Vinolin1 · M. Sucharitha2 Accepted: 5 October 2020 © Springer-Verlag GmbH Germany, part of Springer Nature 2020
Abstract Video forgery detection is one of the challenges in this digital era, where the focus is on discovering authenticity. Though there are so many methods available to detect forgeries in the video, there is no method that utilizes illumination-based forgery detection. Hence, this research focuses on establishing the 3D model of the video frame to generate light coefficients in order to detect the forgeries in the video. On the other hand, this paper proposes dual adaptive-Taylor-rider optimization algorithmbased deep convolutional neural network (DA-Taylor-ROA-based DCNN) for video forgery detection, where DCNN is trained using the dual adaptive-Taylor-rider optimization algorithm (DA-TROA) that inherits the adaptive concept and Taylor series within the standard rider optimization algorithm (ROA). For the detection process, the distance-based features from the light coefficients and face objects detected using the Viola–Jones algorithm from the video frames are used. The significance of the method is analyzed using the real images for varying noise conditions based on the performance metrics, such as accuracy, true positive rate, and true negative rate. The percentage improvement of accuracy for proposed DA-Taylor-ROA-based DCNN with respect to Taylor-ROA-Based deep CNN is 4.3626% in the absence of noise, and 1.5985% of accuracy improvement in the presence of speckle noise, respectively. Keywords Video forgery detection · Deep learning network · Rider optimization algorithm · Taylor series · Spliced images
1 Introduction The digital forgery is increasing in day-to-day lives due to the innovation of several sophisticated image processing tools. One of the familiar forgery techniques is a copy–move forgery, wherein the component of an image is copied to other places in the same image with the purpose of hiding or totaling some content of the image. The goal of the digital image forensics is to discover authenticity on the basis of digital images. The most frequent forgery techniques are copy–move forgery, wherein the part of an image is copied to other locations. The copied region is converted prior to the translation by adapting rotation, scaling, distortion, or amalgamation of these transformations. The goal of forgery is to hide or insert some content on the images. It is a fact
B
V. Vinolin [email protected]
1
Noorul Islam Centre for Higher Education, Kumaracoil, Kanyakumari, Tamilnadu, India
2
Malla Reddy College of Engineering and Technology, Maisammaguda, Medchal, Telangana, India
that the regions that are forged come from the same images, and thus, it becomes impossible to utilize some statistical properties like the noise of camera, illumination conditions, and so on for detecting the forgery as they are well matched [1]. The suspicious manipulation and tamper
Data Loading...
