Local correlation pattern for image steganalysis

摘要: Correlation of pixels is the most important information used for image steganalysis. Current methods often consider some special types relationships among neighboring pixels. In this paper, we propose a general descriptor to correlation comprehensively. We in an adjacency pattern as local (LCP). The LCP proposed embrace different patterns and represent each by mapping relative values numerical value. Then, histograms are taken features also can be describing elements residual obtained filtering. Experiments show that our constructed feature set based on outperforms state-of-the-art method detecting three popular steganographic algorithms.