LAUSR

Steganography aims to conceal secret data into common media, and steganalysis takes the adversarial position by trying to reveal embedding traces. Most of the existing steganographic schemes for spatial images are designed for gray-scale images. Their extensions to color images are often simply processed by treating each color channel as a single gray image and distributing payload uniformly to different color channels. However, color images are commonly used in daily life, and some effective steganalytic methods recently proposed for color images have improved detection performance with features designed by taking color correlations into consideration. In this paper, we propose a novel steganographic scheme for spatial color images by considering color pixel vectors (CPVs), which are composed of color components in the same spatial location, as embedding units. Embedding costs are directly defined on CPVs, and therefore, intricate relationships among color channels can be explicitly considered, and embedding payload is adaptively assigned to all three color channels. To further enhance the performance, an adapted clustering modification directions strategy is incorporated. The experimental results show that the proposed scheme can effectively resist the steganalytic methods, especially with color-rich model features.