| Digital video as the main information carrier is widely used in our daily work and life.However,the advent of video editing software,which is fully functional and easy to use,has made it simple and interesting to make changes to digital video.Although some people edit the video just to enhance the video performance,but there are some people for malicious attacks.The undistinguishable digital video gives a chance to the criminals,the digital video forensics technology has emerged as an important technical means to verify the authenticity and integrity of video in recent years.As tampering with video necessarily requires the process of recompression,double digital video compression forensics technology can be used as the first line of defense to detect the authenticity and reliability of the video.As the latest generation of coding standards,HE VC has a very wide range of application prospect,so it is of great practical significance to carry out double HEVC compression detection algorithms.Based on the deep research of HEVC core technology and key innovative technologies,the changes of HEVC video characteristics caused by recompression is analyzed in this thesis.Based on the unique features and the common characteristics between HEVC and other coding standards,a feasible and innovative double digital video compression detection algorithm with the same quantization parameters and different bit rates is present in this thesis.The main work of this thesis is as follows:(1)Double HEVC compression detection algorithm based on the PU numbers with the same Quantization parameters is proposed in this thesis to identify double HEVC compression videos.Firstly,the number of PU types with the size of 4×4 in ?-frames of all the GOP is extracted in the double encoding process,and then the standard deviation(SDoPU)of the 4×4 PU types numbers before and after recompression with the same Quantitative parameters is calculated.In addition,record and analyze the SDoPU characteristic curve of the singly compressed video and recompressed video,and finally select an appropriate threshold to achieve the classification of singly compressed video and recompressed video.The experimental results show that the average detection rate of the algorithm reaches more than 80%,which can effectively identify double HEVC compression video,and further promote the study of double HEVC compression forensics.(2)Double HEVC compression detection algorithm based on the Co-occurrence matrix of DCT coefficients and PU types with different bit rates is proposed in this thesis to identify double HEVC compression videos.Firstly,the 5×5 co-occurrence matrixes of DCT coefficients are extracted respectively from the horizontal,vertical,major and minor directions,and then combined with the 4×4co-occurrence matrices of PU types in the four directions to construct a 164 dimensional joint feature,which is then be sent to SVM to classify singly compressed videos and recompressed videos.The experimental results show that the detection rate of the proposed algorithm is 100%,which proves that the proposed algorithm achieves a good classification effect for HEVC singly compressed video and recompressed video.In addition,this thesis reduces the dimensionality of the joint feature,that is,it only uses the horizontal co-occurrence matrix of DCT coefficients and PU types to construct a 41 dimensional joint feature to identify whether the video has undergone recompression,the detection rate still reaches more than 96%.Finally,robust test of this algorithm is carried out.In the experiments,we constructed the recompressed video datasets by deleting the frames of 10%,30%and 50%respectively,the SVM classification results show that the proposed algorithm has strong robustness to frame deletion,which further verify the effectiveness of the proposed algorithm. |
PDF Full Text Request