ABC optimized adaptive fuzzy PVD steganography

This paper presents an adaptive image steganography method that combines fuzzy-based capacity selection with a Preserved-Population Artificial Bee Colony (PP-ABC) optimization approach. The method operates on nonoverlapping 2×2 image blocks, where the top left pixel is kept as a reference pixel and the remaining three pixels are used for secret data embedding. Local neighborhood differences between reference pixels determine adaptive embedding capacities through percentile-based fuzzy membership functions. For each embeddable pixel, feasible 8 bit pixel values satisfying the corresponding modulus and boundary conditions are created and preserved. The PP-ABC searches for low distortion pixel solutions while preserving feasible population members. A 32 bit payload length header supports blind recovery of the embedded secret data. The method was evaluated on nine 512×512 grayscale images from the USC-SIPI image database. Embedding capacities range from 654,546 to 670,021 bits, with an average of 660,911 bits. Under controlled payload comparison, PP-ABC achieves an average MSE of 20.6849 and PSNR of 34.9756 dB, with NCC, SSIM, and QI values of 0.991266, 0.9211, and 0.991221, respectively. Payload rate experiments show decreasing PSNR with increasing embedding rate. RS steganalysis and pixel difference histogram analysis characterize the statistical changes introduced by embedding.