Research On Image Copy-Paste Detection Based On Quantum Clustering

Image copy-paste tampering can cause serious adverse effects on social development and personal credibility.Therefore,image copy-paste tampering detection research is of great importance for strengthening social credibility,maintaining judicial justice,and protecting national information content security.With the increasing maturity of image processing technology,traditional image copy-paste tampering detection methods have revealed many limitations.On the one hand,due to the increasing image quality,the complexity of image key feature-based detection algorithms increases;on the other hand,due to the increasing complexity of image postprocessing operations,more and more features need to be extracted in the image detection process,consuming a large amount of computational resources.Therefore,the thesis takes advantage of the superposition and entanglement properties possessed by quanta and the parallelism of quantum computing to study quantum clustering algorithms to solve the problems of low efficiency and computational complexity in the process of matching key point features of tampered images.The main research work can be summarized into two aspects as follows1.A quantum particle swarm clustering algorithm based image copy-paste detection method is explored for the problem that the number of key points targeted in the tampered image copy-paste detection process is large and the traditional clustering algorithm is easy to be locally optimal leading to inefficient matching of feature points in image copy-paste detection.The method first extracts the feature key point vector by SIFT algorithm;then clusters the feature key point vector using quantum particle swarm algorithm and performs key point matching in the clusters;finally removes the wrong matching points according to RANSAC algorithm and detects the image copy-paste tampering matching point pairs.The experimental results show that the method with improves the accuracy of image copy-paste detection to a certain extent and can also show good performance for post-processed images.2.A copy-paste detection method based on quantum k-means clustering algorithm is explored for the problems of high dimensionality of image copy-paste detection feature vectors and inefficient feature matching.The method first extracts the image key point feature sub-description vector according to SIFT and A-KAZE;then the feature point description vector is quantitatively processed by You-transform and blackbox Oracle operation,and the clustering center is selected by quantum mode distance,and quantum similarity finding is used for clustering;finally,the image copy-paste region is obtained by feature matching of the clustered feature key points.The experimental results show that the method reduces the time consumption in image detection and still shows good robustness for post-processed images.