Image re-sampling detection based on Markov process and Gabor filtering

Abstract

An Image re-sampling tampering detection algorithm is one of the important methods for image tampering evidence collection. According to the invention, an image is detected mainly based on condition state transition probability matrixes in a Markov process and the texture features in a Gabor filter. After an image is re-sampled, the correlation between adjacent DCT coefficients changes. However, the transition probability matrixes can well describe the change in DCT coefficients and surrounding coefficients. Thus, transition probability matrixes in four directions are extracted as Markov features. The detection rate of the algorithm fluctuates greatly when comes to texture images of different complexity. Thus, a Gabor filter group is adopted to extract the texture features of the image and perform two-dimensional discrete Fourier transform to get the final texture features of the image, and then, an SVM is adopted to train and classify the two types of features. With the algorithm of the invention, scalable images and rotary images can be well detected, and detection of rotary images is of high robustness.

Description

technical field [0001] The invention relates to an image resampling tampering detection method combining statistical features and texture features, and belongs to the technical field of digital image tampering evidence collection. Background technique [0002] In today's era of rapid development of digital information, digital images, as the most commonly used digital information carrier, are widely used in news, justice, science, entertainment, military and other fields. Sexuality has caused serious damage, hindered judicial justice, and affected social stability and order. Therefore, digital image forensics technology has become a very important direction in the field of information security technology research. [0003] There are many kinds of image tampering operations, such as image stitching, paste-copy, resampling and other operations. The present invention mainly detects for resampling operation, and the detection of resampling can be divided into two types: People...

AI Technical Summary

Problems solved by technology

But every kind of method all has their shortcoming, and main shortcoming has the robustness of algorithm not strong, promptly can not take into account simultaneously to rotation and scaling operation, or the calculation of scaling factor etc. in the image, so in the present invention, aim at algorithm Improve some shortcomings of the algorithm to improve the detection accuracy of the algorithm

Images