Reversible Data Hiding for Clustering Based on Pixel Texture Features

In the big data era, information security has gained growing importance. In this paper, a reversible data hiding method using DBSCAN clustering is introduced. Firstly, the initial image is partitioned into four distinct regions using a grid-based check-erboard segmentation pattern. Subsequently, one of the regions is selected as the target region, with the pixels within this region serving as the objects for data embedding. The corresponding feature vector is constructed using the pixels from the remaining three regions for each pixel in the target region. The feature vector is composed of the contextual information of the target pixel, which effectively captures the changing trends of the local structure within the image. Subsequently, we utilize the feature vectors to perform DBSCAN clustering within the target region. Finally, based on the clustering results, data embedding is performed in each cluster with a 1×q block size. Experimental findings indicate that the proposed method offers notable improvements compared to current state-of-the-art techniques.