Tampering Detection Of Seam Carving Based On Orthogonal Polynomial Transform

With the rapid economic growth in recent years,digital image plays an increasingly important role in the development of the times as an important way of information dissemination in today’s life and work.Images appear in the form of digitization,which can be easily stored,modified,copied and transmitted.Seam carving is a content-based image scaling technology.Compared with the traditional image scaling technology,image seam carving calculates the energy of each pixel of the target image one by one,and removes the set of pixels with the lowest energy value to retain the high-energy area as much as possible.However,when the tampering ratio is large,the image will still be distorted.At present,the detection algorithms for image seam carving tampering are mainly based on feature extraction,and most of them are for JPEG format image tampering detection.At the same time,when the tampering ratio is small,the detection accuracy is not high.In this thesis,the characteristics of image seam carving are studied deeply.The main work includes the following aspects:1.Since some pixels are deleted by image seam carving,a tamper detection method is proposed based on the idea that all pixels in the image are related.In this method,Discrete Tchebichef Transform(DTT)is used to extract the trace of seam carving tampering,and realize the detection of seam carving tampering.Firstly,the proposed method divides the input image into 8×8 non-overlapping blocks and followed by carrying out the DTT on each block to get the transformed matrix.Secondly,the proposed method calculates the difference between the coefficients in the transformed matrix of each block and gets the histogram of difference,and the statistical matrix is obtained from the difference matrix.Lastly,the features are extracted from the statistical matrix.Experimental results show that this method is not only suitable for JPEG format and TIFF format tampered images,but also can achieve high detection accuracy for small proportion of tampered images.2.In this thesis,a seam carving optimization algorithm based on Hessian matrix is proposed.An edge detection image is obtained by using Hessian decomposition matrix.The edge detection image can more effectively distinguish the high-frequency and lowfrequency regions in the image,and avoid the high-frequency region or the important content passing through the optimal seam.In addition,the method is further optimized when selecting seams.Among the 10 seams with the lowest cumulative energy in the cumulative energy map,the seam with the worst physiological visual effect is selected to remove.In this way,the content of the image is not damaged,and the good human physiological visual effect is maintained.Experimental results show that the algorithm can effectively achieve the purpose of image scaling.