Computer Generated Colorized Image Forgery Detection using VLAD Encoding and SVM
Ze Yang, Zhenyu Yu, Yuying Liang, Rui Guo, Zhihua Xiang · 2020
With the development of multimedia technology, more and more applications used in digital image processing, tampering with digital images is become easier. This paper proposes a new approach for image detection based on color tampering by encoding multiple color channels features based on VLAD without embedding watermarks. We have counted multiple sets of common color channels in computer vision, choosing the most suitable combination of color channels, and then using VLAD to encode these selected features, finally to train a SVM model by encoded features, we also found that when a lot of classes image were mixed together cannot be will detected, so we adopted deep learning to train a ResNet of classification, and first to classify our dataset, in comparison with state-of-theart method, various experiments result prove that the proposed approach achieves better performance in computer generated colorized image forgery detection.