Apparatus and method for steganalysis
AssigneeNew Jersey Institute of Technology
InventorsYun-Qing Shi, Guorong Xuan
Life of this patent
Estimating the term…
Abstract
An apparatus and method for steganalysis that enhances the ability to detect distortion introduced by data hiding. In embodiments of the invention, a pixel grayscale value in an image is predicted by using its neighboring grayscale values of neighboring pixels. Further, a prediction-error image is produced by subtracting the image from its predicted image. The prediction-error image may is employed to remove at least some variations in image data other than those associated with data hiding an thus, at least partially offsets variations from image aspects other than data hiding.
Claims18
An apparatus comprising:
- means for generating features based at least in part on moments of a characteristic function of an image and said image with a watermark embedded;
- and
- means for classifying said image and said watermark embedded image based at least in part on said generated features.
An apparatus comprising:
- means for applying a trained classifier to an image;
- means for applying a tested classifier to a watermark embedded image;
- and
- means for classifying said image based at least in part on applying a trained classifier to a host of features generated from said image and said watermark embedded image.
A method for steganalysis of an image, comprising:
- generating a prediction-error image from a gray-scale of the image;
- computing discrete wavelet transforms (DWTs) of the gray-scale image and the prediction-error image;
- computing histograms of the gray-scale image and the prediction-error image, designating each as an LL0 sub-band of each image;
- computing histograms of LLi LHi, HLi and HHi sub-bands, where i=1, 2, 3, of the DWTs of the gray-scale image and the prediction-error image;
- computing moments for each of the sub-bands;
- combining the moments to extract features from the gray-scale image and the prediction-error image;
- and
- analyzing the extracted features with classifiers configured to detect changes in the histograms that indicate hidden data.
Dependencies are inferred from the claim text.