Research On Key Identifying Technology Of Photorealistic Computer Graphics And Recaptured Image

With the rapid development of the Internet,it is inevitable to transmit the image in such a large digital network.But these images can be copied,modified and added easily.The criminals can use these created images to cause malicious damage.Although it can be judged whether the image is modified by related technology,with the rapid development of digital cameras and smart phones,there are some new ways of image attacking mode generated that are computer-generated image and recaptured image.The high-quality forgery image has brought us a great threat.For example the South China tiger incident occurred in Shaanxi,China in 2007 has prompted us that the authenticity of digital images is not only limited to the tampered image and also not necessarily true for the real image from digital camera and smart phone.Therefore,it is significantly important to identify the real image and the forgery image.The main work:1.The corresponding mathematical model is given according to the image processing methods.The existing image authentication algorithms are summarized and their own views are put forward.The difference of between the real image,computer generated image and recaptured image is put forward.A mathematical model for imaging difference between real image and recaptured image is proposed.The difference between high frequency information and low frequency information of real image and remake image in frequency domain is put forward.2.The computer generated image forensics algorithm based on sub-pixel and LTP is proposed.The algorithm uses the sub-pixel principle and the Hough line detection algorithm to extract the edge linear information of the image.Then the differential sub-pixel feature,gradient direction sub-pixel feature and gradient variance sub pixel feature are extracted,and these features are as the final feature vector for classification.At the same time,we use texture algorithm to extract image features.In this paper,the sub-pixel feature and texture feature are extracted as the final feature vector for classification.The experimental results show that the algorithm has a good forensics rate for the computer generated image.3.The recaptured image forensics algorithm based on Local plane linear point is proposed.In this paper,the proposed algorithm establishes the mathematical model in the imaging process and provides concepts and properties of the local plane linear point from the model.Then the local plane linear point was extracted from image as the characteristic value.Finally the support vector machine is applied to classify the recaptured image with the characteristic value.4.The recaptured image detection algorithm based on co-occurrence matrix is proposed.According to the difference between the real image and recaptured image in the image texture,the wavelet transform is used to separate the low frequency information and the high frequency information of the image.Then horizontal,vertical,and diagonal co-occurrence matrices of low frequency images and high frequency images are got.The contrast feature,energy feature,entropy feature and correlation feature of co-occurrence matrix are extracted as the feature vectors.Compared with other similar algorithms,the proposed algorithm in this paper has excellent recognition rate for recaptured images with background.5.The recaptured image identifying algorithm based on texture feature is proposed.The difference in image is used to propose the algorithm based on the light source chromaticity,the noise image and steerable pyramid.For the difference between high frequency image and low frequency image,the steerable pyramid is used to extract the low frequency features,high frequency images features and directional features from the image.For the noise difference,the Gauss filtering algorithm is used to extract the image noise characteristics which are used to analyze images.For the difference between the real image and the recaptured image in the light source information,the chromaticity features of light source are extracted.The results show that the proposed algorithm has good recognition rate for both the background images and the background images without background information.