The Copy-Move Forgery Detection Based On RGB Color Features And Discrete Wavelet Transformation

With the increasingly sophisticated development of digital technology and image editing software being widely used, ordinary computer users can easily change the content of images to generate images as real. To some extent, these software can bring us convenience and enjoyment, however, if used improperly, they will bring very serious consequences even affect the military, political, diplomatic capability of a Country. Therefore, the research for digital image forensics is very indispensable. The steady improved digital image forensics provides a guarantee of the image content's reality and security at the same time it can bring us convenience and entertainment.Digital image blind forensics technology was proposed for the forgery of image content, and as one important aspect in detecting methods, copy-move detection has been researched deeply by lots of scholars. The forgery of copy-move is one important tamper mean including copy-move in one image and between different images. In this paper, we focus on copy-move in one image with detecting if there are two same or similar regions or not. Currently, detection methods include traversal research, image block autocorrelation matrix and image block matching. Although detection results of the above methods are very good, it has so large computation and more dimensions of feature vectors that the running time of program is too long.In this paper, we combine discrete wavelet transform (DWT) with RGB color features to resolve the above problems. Firstly, we extract the RGB components of the image to be detected on which we do DWT to obtain the low-frequency sub-band of the original image. Secondly, we break the low-frequency sub-band up into many small overlapping blocks with size a*a. We assume the size of the original image is M*N*P, then the size of the RGB components are m*n, and the number of overlapped blocks is (m-a+l)*(n-a+1). We extract 7 vectors from every block as block feature vectors and store these vectors in the matrix. Finally, we can look for the similar blocks pair through determination conditions.