Digital Image Forensics Technology Based On The Color Characteristics

ABSTRACT:With the rapid development of multimedia information technology, digital images are applied in many fields. The image processing technology also have developed rapidly, and the image manipulation have become more easier. Most of the forged images are hard to distinguish with the human eye. And this phenomenon also creates the trust crisis in media, politics and academia. It makes the public doubt the authenticity of images. So it is very important to detect the authenticity of digital images.this paper introduces the present image detection technology, and analyzes the digital imaging process, principles and features of forensics technology. Meanwhile this paper also describes the general implementation framework of digital image blind forensics process.Based on the summary of existing digital image forensics technology, this paper introduces the influence on color components by reflection model of object surface and digital imaging process. And it describes the variation of pixel values in the edge transition zone in image. Based on feature classification process, digital image tampering detection algorithm using the support vector machine(SVM) is designed in this paper. The one method makes use of the high-frequency characteristics of edge in color difference planes, and another method uses the edge transition region characteristics of color difference planes.The image forensics technology based on edge high-frequency feature utilizes the correlation among natural image three color components. Method analyzes the correlation in spatial domain and frequency domain, and further analyzes the relation among high frequency in different color planes. Then the technique identifies the authenticity of image using the high-frequency characteristics of color difference planes.The main idea of other method is about the edge transition area of color difference planes. The natural images have edge transition feature because of the imaging system. The ideal edges do not exist in natural images. And the color difference planes are more smooth. We abstract the characteristic values of transition region as classification features to detect the authenticity of images.Finally, a summary of our work is summarized. Furthermore, the future work of the thesis is presented in paper.