Blind Detection Algorithm For Image Smoothing Filter

With the popularization of computer, various image carriers are everywhere. The development of digital image processing technology also becomes rapidly.How to judge the authenticity and integrity of acquired image has become a hot research topic.Blind image forensics without relying on pre-embedded information has become a hot issue in the field of information security techniques.Based on digital image filtering blind detection algorithm is one of the emerging research.In the process of image tampering and forgery, different degrees of filtering on image is conducted in order to hide tampering traces and interfere image detail information.These malicious operations make image information's reliability and authenticity descend. Therefore,to determine whether the image is filtered or not is not only very important but also the prerequisite of image forensics and the study of image tampering.According to the properties of typical smoothing filter, this article conducts deep study in frequency domain. This article is mainly embodied in two aspects:(1)The filtering operation can not only suppress image noise effectively.Some filter can also protect details and high frequency component of images, which is mainly due to the impact of different filters with different cut-off frequencies.In this paper, a detailed study on spectrum of the filtered image is conducted, and experiments show that the frequency characteristics of filters is different in the frequency domain,especially in the direction. Through the analysis of frequency domain characteristics of different parameters in filter, the relationship can be obtained between filter parameters and cut-off frequency, and applied in the detection of filter.(2)This paper proposes a detection algorithm for linear filter. By the method of coding image spectrum of testing image, extracting local feature of code image with local xor operator LXP(Local Xor Pattern) and weight algorithm, center of image features of circular radius size can be measured.In this way to judge whether the image is suffered filtering operation and the corresponding filter parameters by matching the circular radius size with the cut-off frequency.