Two-pass hashing feature representation and searching method for copy-move forgery detection
| Autor: | Jun-Liu Zhong, Chi-Man Pun |
|---|---|
| Rok vydání: | 2020 |
| Předmět: |
Sequence
Information Systems and Management Computer science business.industry 05 social sciences Hash function 050301 education Pattern recognition 02 engineering and technology Computer Science Applications Theoretical Computer Science Transformation (function) Artificial Intelligence Control and Systems Engineering Feature (computer vision) Search algorithm 0202 electrical engineering electronic engineering information engineering 020201 artificial intelligence & image processing Artificial intelligence Representation (mathematics) business 0503 education Software Block (data storage) |
| Zdroj: | Information Sciences. 512:675-692 |
| ISSN: | 0020-0255 |
| DOI: | 10.1016/j.ins.2019.09.085 |
| Popis: | In this paper, we propose a two-pass hashing feature representation and searching method for copy-move forgery detection with a good score and high efficiency. First, the normalized moment transformation is presented to extract the corresponding block features from multiple frequency images. The multiple-dimensional features of each pixel are projected into the corresponding hashing bin to obtain the corresponding hashing features. Then, a novel two-pass hashing feature representation is proposed to concatenate multiple hashing features as the bit sequence. Based on the two-pass hashing feature representations, the two-pass hashing searching algorithm searches and updates the nearest pixel matches in high efficiency. Finally, post-processing operations are proposed to accurately identify the forgery regions. The experimental results show that the proposed copy-move forgery detection method can achieve the best scores among the state-of-the-art methods, even under various attacks. In addition, the proposed method has a very high detection efficiency without iterations. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|
Full Text from ScienceDirect