A Study And Implementation Of Image Splicing Detecting Algorithm Based On Camera Instinct Parameter

As the development of the digital photography technology and the popularity of digital camera, digital images play a more and more important role in daily life. On the other hand, the excellent imaging processing software, such as Adobe Photoshop, makes the image modification much easier than before. And brilliant forgeries technology hardly leaves any clues. Thus we are unable to tell the authenticity of an image only by eyes.There are many kinds of images, with a tremendous difference of content. So image authenticity also needs multitudinous tools. So far, there are many approaches to authenticate image. But none of them are effective for all kinds of images and forgeries technologies. Two or more images splicing is consistently used, which is the most practical forgeries technology.This article aims at photos of two or more people and images with abundant geometry information. Based on theories of computer vision and projective geometry, we propose a new approach of estimating camera intrinsic parameters and then investigating the rationality of camera instinct parameters to authenticate image splicing. With regard to images of people, the algorithm extracts the matrix of the camel sclera rims. Then, the coordinates of circular points are respectively calculated, and intrinsic parameters of the camera are estimated using circular points. For the images with rich geometric conditions, the algorithm takes plane as a unit, extracts information from every plane, such as coordinates of points on the plane, line equations, a rectangle with known length and width, or quadratic equations.According to calculations of the points, lines, rectangles, or the second curve on each plane, each plane's homography between the positive view of the real object and the image are estimated. Then according to homography, we modify the Zhang Zhengyou's plane calibration algorithm to solve the intrinsic parameters of the cameras. For each image with the parameters of the cameras, we inspect the rationality and consistency of the principal points and the focal to decide whether the image is splicing together.The main contribution of this article is, a new approach to image splicing probability detection, using the camera intrinsic parameters. It examines the authenticity of images from the perspective of computer vision. This article breaks the existing camera calibration algorithm's restrictions of demanding multiple images and the specific calibration model. It modifies the existing classical camera calibration algorithm, successfully using camera parameter estimation for passive authentication of splicing images, and has made good results.