Image Steganography Based on Pyramid Pooling Module

Abstract

In this paper, we add a pyramid pooling module to the depth model of image steganography, which can effectivelyimprove the visual effect of steganographic images. The goal of the study is to fully integrate previously important globalfeatures to achieve good hiding and extraction effects and reduce information loss while ensuring security effects. Theexperiments are conducted by randomly selecting images from the ImageNet dataset for training and testing the effect ondifferent datasets. The experimental results show that the peak signal-to-noise ratio (PSNR) and structural similarity ratio(SSIM) between images obtained by this method can obtain good values, which brings further visual improvement for imagesteganography tasks.