Research On Image Copy Tamper Detection Algorithm Based On SIFT Feature Points

Copy and paste is one of the most common ways of image tampering.Copy tampered image will affect the authenticity of media reports and the impartiality of judicial authentication.The existing images copy tamper detection algorithm has the problems of large positioning tampering area error and low detection accuracy.Therefore,it is of great significance for the research of image copy tamper detection.Aiming at the problem of low detection rate of texture smooth areas of copy tampered images as well as poor robustness of post-processing such as rotation and scale transformation,a copy tampered image detection algorithm based on SIFT feature points is proposed.Taking into account the different texture features of the image in this algorithm,the adaptive SLIC algorithm is used to perform non-overlapping and irregular blocking on the image.Extract SIFT features and extract rotation invariant uniform LBP features on the area to be described in SIFT,establishing an improved feature descriptor.Based on the feature descriptor,the standard Euclidean distance,the Correlation distance and the Hamming distance are calculated separately,besides the g2nn algorithm is improved through multi-distance matching optimization to execute the initial feature matching.Use condensed hierarchical clustering and RANSAC algorithm to remove existing mismatches and determine whether the image has been tampered.For copy tampered images,the super pixels are used to determine the suspicious area to replace the marked feature points.Apply variable neighborhood optimization to improve the local search algorithm to determine the neighborhood block,moreover,compare the similarity between the suspected area and the neighborhood block.Combine the suspected area and the neighboring block area to perform morphological closed operation to complete the positioning of the tampered area.In order to visualize the improved algorithm,a GUI-based image copy tamper detection simulation system is designed.Based on the simulation analysis of the two image sets of MICC-F600 and Dataset,the effectiveness of the detection algorithm is verified.The detection accuracy of the improved algorithm is 94.24%,compared with the SIFT algorithm and feature point iteration algorithm,the detection accuracy is improved by 5.35%and 2.10%respectively.The simulation results show that the improved algorithm has high detection accuracy for the texture rich and smooth areas of the copy tampered image and is very robust to post-processing of scale transformation,rotation,blur,noise,and compression.