Research on Reversible Data Hiding Algorithm for Images Based on K-Means Clustering and PVO Adaptive Double-Layer Embedding

Reversible Data Hiding (RDH), as a crucial information security technology, enables the concealment of data while allowing for the lossless recovery of the original carrier and the extraction of secret data. This paper proposes a PVO-based adaptive double-layer embedding algorithm utilizing K-means clustering. The algorithm preprocesses image pixels using K-means clustering to dynamically identify smooth regions and selects optimal embedding regions by combining local complexity thresholds, thereby increasing embedding capacity and reducing image distortion. Additionally, the algorithm employs a double-layer embedding strategy, embedding information at both the maximum and minimum pixel positions within blocks, further enhancing embedding efficiency Experimental results demonstrate that, compared to traditional PVO algorithms, the proposed algorithm achieves an average improvement in peak signal-to-noise ratio (PSNR) of approximately 1.5 dB and 3.3 dB under the same conditions for embedding capacities of 10,000 bits and 20,000 bits, respectively.