Robust and fast image hashing with two-dimensional PCA
- PDF / 1,698,363 Bytes
- 13 Pages / 595.276 x 790.866 pts Page_size
- 58 Downloads / 235 Views
SPECIAL ISSUE PAPER
Robust and fast image hashing with two‑dimensional PCA Xiaoping Liang1 · Zhenjun Tang1 · Xiaolan Xie2 · Jingli Wu1 · Xianquan Zhang1 Received: 9 June 2020 / Accepted: 20 September 2020 © Springer-Verlag GmbH Germany, part of Springer Nature 2020
Abstract Image hashing is a useful technology of many multimedia systems, such as image retrieval, image copy detection, multimedia forensics and image authentication. Most of the existing hashing algorithms do not reach a good classification between robustness and discrimination and some hashing algorithms based on dimensionality reduction have high computational cost. To solve these problems, we propose a robust and fast image hashing based on two-dimensional (2D) principal component analysis (PCA) and saliency map. The saliency map determined by a visual attention model called LC (luminance contrast) method can ensure good robustness of our hashing. Since 2D PCA is a fast and efficient technique of dimensionality reduction, the use of 2D PCA helps to learn a compact and discriminative code and provide a fast speed of our hashing. Extensive experiments are carried out to validate the performances of our hashing. Classification comparison shows that our hashing is better than some state-of-the-art algorithms. Computational time comparison illustrates that our hashing outperforms some compared algorithms based on dimensionality reduction. Keywords Image hashing · Principal component analysis (PCA) · Two-dimensional PCA · Visual attention model · Saliency map · Dimensionality reduction
1 Introduction Image hashing [1, 2] is an efficient multimedia technology. Its input is an image and its output is a compact representation based on the visual content of input image. Generally, the compact representation is called an image hash. In practice, image hash is used to represent its original image. As the size of the image hash is much smaller than that of image data, the use of image hashing technology can achieve efficient processing. Currently, image hashing technology has been widely used in many multimedia systems [3–11], such as image retrieval, digital watermarking, image copy detection, multimedia forensics, image encryption and image authentication. In this paper, we study a new image hashing
* Zhenjun Tang [email protected]; [email protected] 1
Guangxi Key Lab of Multi‑Source Information Mining and Security, and Department of Computer Science, Guangxi Normal University, No.15 Yucai Road, Guilin 541004, China
College of Information Science and Engineering, Guilin University of Technology, Guilin 541006, China
2
based on two-dimensional (2D) principal component analysis (PCA). In general, image hashing should satisfy two basic properties [12, 13]: robustness and discrimination. Robustness refers to the property that the hashing algorithm should learn the same or similar hashes from those visually similar images, which may undergo common digital operations, such as image compression and geometric transforms. Discrimination refers to the
Data Loading...
