Fast Image Clustering Based on Camera Fingerprint Ordering

This work presents a new camera fingerprint-based image clustering algorithm. The proposed algorithm is based on sorting the camera fingerprints according to information that is inherently present in images. A ranking index is constructed for each image, taking into account the combined effect of gray-level, saturation and texture on camera fingerprint estimation. Then, camera fingerprints are ordered according to this ranking index and clusters are iteratively constructed using as reference fingerprint the top-ranked fingerprint among the currently un-clustered fingerprints. The algorithm can be optionally implemented with an additional attraction stage to refine clustering. The results confirm that the proposed method achieves a performance comparable to state of the art approaches, with a significantly lower computational complexity. The method can also handle cases in which the number of clusters is much larger than the average size of the clusters.