Wavelet-Based Multiresolution Features for Detecting Duplications in Images

摘要: Duplication of image regions is a common method for manipulating original images using typical software like Adobe Photoshop. In this study, we propose wavelet based feature representation scheme detecting duplicated in images. This technique works by first applying multi-resolution decomposition to small fixed-sized blocks. Normalized coefficients are then stacked into vector an order from lower higher frequencies. kind appears robust block matching. Duplicated detected lexicographically sorting all the blocks and threshold desired frequency offsets blockcoordinates. A semi-automatic that detects accurate number also proposed. Initial experiments with set natural having show impressive results compared linear PCA representation.