Image Approximate Copy Copyright Detection Technology and Algorithm for Network Propagation Under Big Data Condition

摘要

This paper adopts the viewpoint of image copy detection which combines local and global features, gives full play to the matching advantages of different image features, dynamically adjusts the weight of feature fusion according to the expressive ability of different image features and the image content characteristics of different image data sets, calculates the optimal weight allocation value, and proposes the dynamic update weight allocation calculation. This method further improves the accuracy and comprehensiveness of image approximate copy detection.