Perceptual Color Image Hashing Based on Quaternionic Local Ranking Binary Pattern.

Image hashing is a technique that represents an image as a short unidirectional sequence. In this study, we propose a robust color image hashing technique based on the quaternionic local ranking binary pattern (QLRBP). First, the values of the RGB color channels are expressed as a pure quaternion, and the texture image is generated using the phase difference of the Clifford translation quaternion as the source for local binary pattern (LBP) comparison data. Next, mean pooling is used to reduce the dimensionality of the image texture. For security, the pixels of the dimension-reduced image are randomly selected, and each pixel is quantized to obtain the final image hash. Experimental results showed that the proposed scheme resists a variety of ordinary image operations and achieves good balance between robustness and discrimination. In comparison with other state-of-the-art methods, the receiver operating characteristic (ROC) curves of our scheme demonstrate its superiority in terms of content identification. [ABSTRACT FROM AUTHOR]