A smoothing filter operation type and parameter identification method

Abstract

A smoothing filter operation type and parameter identification method, comprising: performing discrete Fourier transform on an acquired image to be tested to obtain a spectral response of the same size as the image to be tested, and performing modulo processing on all values ​​in the spectral response to obtain the image to be tested The frequency-domain amplitude map of the image; process the frequency-domain amplitude map to obtain a binarized frequency-domain map that can reflect the frequency-domain distribution characteristics of the image to be tested, and then use statistical methods to extract N-dimensional images from the binarized frequency-domain map Detection features: Input the extracted N-dimensional detection features into the support vector machine for training to obtain a multi-class detector, which is used to detect whether the image to be tested has undergone smoothing filtering, and to identify the type and parameters of the smoothing filtering operation experienced. The invention realizes effective detection and classification of smoothing filtering operations represented by mean value filtering, Gaussian filtering and median filtering through effective characterization of images, and further realizes parameter identification of specific filtering operations. The detection feature dimension of the method of the invention is low, and the detection accuracy rate is high.

Description

technical field [0001] The invention relates to a tamper detection method. In particular, it relates to a smoothing filter operation type and a parameter identification method. Background technique [0002] As digital images are widely used, image editing software such as Photoshop is also becoming more and more popular. This makes tampering with image manipulation much easier. Whether it is in professional fields such as justice and medicine, or in entertainment fields such as photography and fine arts, the authenticity and integrity of images have received more and more attention, which has also brought more and more challenges. In order to ensure the authenticity and integrity of digital information, digital image passive forensics technology came into being. Different from the active forensics technology represented by digital signature and digital watermark, the passive forensics technology is only based on the acquired digital image to find out whether there are var...

AI Technical Summary

Problems solved by technology

The above methods can effectively detect the median filter image, but none of them fully involve the identification of smoothing filter operation types and parameters

Images