DHNet: Double MPEG-4 Compression Detection via Multiple DCT Histograms

In this article, we aim to detect the double compression of MPEG-4, a universal video codec that is built into surveillance systems and shooting devices. Double accompanied by various types manipulation, its traces can be exploited determine whether forgery. To end, present neural network-based approach with discriminant features for capturing peculiar artifacts in discrete cosine transform (DCT) domain caused MPEG-4 compression. By analyzing intra-coding process which performs block-DCT-based quantization, exploit multiple DCT histograms as focus on statistical properties coefficients multiresolution blocks. Furthermore, improve detection performance using vectorized feature quantization table dense layers auxiliary information. Compared approaches suitable exploring subtle manipulations, experimental results reveal work achieves high performance.