Video Forgery Detection Based On Macroblock Type Features

In recent years,with the rapid development of Internet technology and transmission technology,the digital video has become one of the most important ways to acquire information in our daily life.However,due to the improvement of the video editing software,the digital video can be easily modified.The integrity and authenticity of the digital video cannot be guaranteed.Hence,in the field of passive authentication,it is of great importance to research how to guarantee the authenticity of the digital video content.In this paper,we propose two passive authentication algorithms of double compression detection.We present an algorithm based on the dual encoding parameter model to detect double H.264 compression with the same quantization parameter.The model is composed of Intra Prediction Macro Block Modes(IPMBM)and the quantized DCT coefficients of I frames.The proposed algorithm designs the classfication feature based on the property that coding parameters of single compressed videos and double compressed videos have different convergencies.The number of altered IPMBMs and the quantized DCT cofficients between two adjacent compressed videos is monotonically decreasing with the increasing compression times.At the same time,the decline is becoming slower and slower.According to this principle,the extracted feature fed to the trained support vector machine(SVM)can detect double H.264 compression correctly and efficiently.We present an algorithm based on Intra Prediction Macro Block Modes(IPMBM),the quantized DCT coefficients of I frames and Inter Prediction Macro Block Modes of P frames(PMBM)to detect double H.264 compression with the same quantization parameter.We count the number of altered IPMBMS,the changed quantized cofficients of I frames and the number of altered PMBMS of P frames between two adjacent compressed videos.The numbers used to calculate the classification feature is fed to the trained support vector machine(SVM)to detect double H.264 compression.