Detection Of Double H.264 Compression

With the rapid development of multimedia technology,the price of the camera equipment is becoming lower and lower,and the video editing software makes it easy for people to tamper with the digital images and videos.The tampered images and videos can easily deceive and confuse the public.Video forensics technology can provide the basis for judicial evidence,which has important social significance.Double compression detection is a very important technology of video forensics.Generally the tampered images and videos have undergone at least two times of compression.Double compression detection can provide some reference evidence for the authenticity of the images or videos.With the H.264 compression standard widely used in mobile phones and video cameras,the study of H.264 video forensics technology is very necessary.Most of the proposed methods in the field of double compression detection are based on MPEG-X series standards,and the proposed methods based on H.264 are less.In 2015,the only method for double H.264 compression detection with the same quantization parameter is proposed.However,no method for double H.264 compression detection with the same bit rate has been proposed prior to this point.This paper achieves breakthrough in the field of double H.264 compression detection with the same quantization parameter or bit rate,and fills some gaps.This paper analyses of the impact of double compression to H.264 videos,extracts features in the macroblock level,and proposes three double H.264 compression detection methods,based on inter prediction macroblock mode feature,intra prediction macroblock mode feature,and fusion macroblock mode feature respectively.The work and innovation of this paper include:1.This paper proposes a method for double H.264 compression detection with the same quantization parameter or bit rate,based on intra prediction macroblock mode feature.Intra prediction macroblock mode is composed of intra prediction macroblock type and intra prediction mode.When repeatedly compressed with the same quantization parameter,the number of intra predicted macroblocks whose modes change between the adjacent compressed versions,decreases with the increase of compression times.The feature is extracted from the I frames of H.264 videos and the support vector machine(SVM)is used for training and classification.The experimental results show that the proposed method can achieve high accuracies in the situations of using the same quantization parameter(VBR mode)or using the same bit rate(CBR mode).2.Finally,this paper proposes a method for double H.264 compression detection with the same quantization parameter or bit rate,based on fusion macroblock mode feature.The fusion feature is composed of inter prediction macroblock mode and intra prediction macroblock mode.Experimental results show that the proposed method achieves a higher accuracy in the case of low video quality.