| ABSTRACT:With the rapid and extensive development of image processing technology and tools, it's increasingly easy to tamper images. These forgeries, when used in the mass media or courts of law, can have bad effect on our society. Hence, it is very important to detect the doctored images.The main reason we study the double-compressed features is that the features usually come from digital tampering. This observation could in principle be used to identify manipulated areas in digital images, while double-compressed features can improve the accuracy and reliability of estimated stego analysis.In this rapper we first identify characteristic features that occur in DCT histograms of individual coefficients due to double compression. Then, we present 2 different approaches that estimate the original quantization matrix from double compressed images.In the first method we scale the image by 0.98 to get the approximate original uncompressed image then we simulate the double compression to the pixel of the same location on the image to get the histogram of DCT coefficients. We can get the original quantization step by calculate the difference between histogram of DCT coefficients of the current image and the histogram of DCT coefficients we get above based on the regularity of the histogram ofDCT coefficients.The second method is based on Method One. Considering 3 different situations of quantization step in the double compression process,we put forward the corresponding solutions. By making the estimation accuracy of the original quantization table greatly improved and the adaptability enhanced.Then, the programming implementation which aimes at the above algorithm is given, and an estimation system of Primary Quantization Matrix in Double Compressed JPEG Images is completed, Experimental results show that the system's output is in line with the expected results.Finally, a summary of the work is given, what's more, we expect the future work related to this thesis. |
PDF Full Text Request