Blind Image Clustering Based on PRNU With Reduced Computational Complexity

The blind clustering of images without any prior information is always challenging for forensics experts. The problem gets severe in the case of a large dataset where the computational cost increases significantly. The proposed technique is one of the potential solutions for image clustering, especially for large-scale clustering. The technique clusters images based on camera fingerprints using two-stage clustering and additional processes. The algorithm is tested on the sub-datasets of the well-known Dresden dataset. The performance of the clustering framework is either better or comparable with the state-of-the-art clustering algorithms. The algorithm generates good-quality clusters with a significantly lower computational complexity.