Image steganography detection method and system based on nonlinear hybrid kernel feature mapping

The invention relates to an image steganography detection method and system based on nonlinear hybrid kernel feature mapping. The method comprises the steps: obtaining a to-be-detected image, and carrying out feature extraction on the to-be-detected image by adopting a rich model type feature extraction method; determining statistical characteristics of the to-be-tested image; statistical featuresof a to-be-tested image are longitudinally segmented in sequence; obtaining a statistical feature block set of the to-be-tested image; mapping each to-be-tested image statistical feature block in theto-be-tested image statistical feature block set by utilizing a feature mapping algorithm based on a nonlinear mixed kernel function; obtaining a statistical feature block set of the mapped to-be-tested image; splicing each mapped to-be-tested image statistical feature block in the mapped to-be-tested image statistical feature block set, synthesizing high-dimensional rich model features of the to-be-tested image, and dete