Group Flamingo search PyramidMobileNet for pixel prediction in image steganography

Protecting medical information during the process of image steganography has received much research interest. Among the available methods, the Least Significant Bit (LSB) substitution is leveraged for hiding essential information. Here, each pixel was changed to enhance privacy; however, it increased the complexity. To overcome this issue, a potent approach named Group Flamingo Search Algorithm_PyramidMobileNet (GFSA_PMNet) was developed for pixel prediction in image steganography. This process consists of two stages: the embedding phase and the extraction phase. The process begins with the embedding phase, which is then followed by the extraction phase. The input medical image is first provided to the feature extraction stage. Then, pixel prediction is performed by utilizing a hybrid PMNet, which is the combination of PyramidNet and MobileNet. PMNet is trained with GFSA, a hybrid of the Group Search Optimizer and Flamingo Search Algorithm. Next, the predicted pixel undergoes the Morlet wavelet transform, after which the secret image is embedded into the predicted pixel. In the final step, the inverse Morlet wavelet transform is used to reconstruct the embedded image. Morlet wavelet extracts features; selected regions reconstruct hidden message, while a CNN performs steganalysis by detecting embedded patterns in images. Moreover, the GFSA_PMNet attained the greatest correlation coefficient of 0.798, Structural Similarity Index Measure (SSIM) of 0.979, Peak Signal-to-Noise Ratio (PSNR) of 41.87 dB, Feature Similarity Index Measure (FSIM) of 0.978, Visual Information Fidelity (VIF) of 0.968, Universal Quality Index (UQI) of 0.867 and Learned Perceptual Image Patch Similarity (LPIPS) of 0.155.