GENERAL VIDEO STEGANALYSIS METHOD BASED ON VIDEO PIXEL SPACE-TIME RELEVANCE

摘要

Disclosed is a general video steganalysis method based on a video pixel space-time correlation. The method comprises: decompressing a video in a training set and extracting a plurality of image sections, and slicing each image section to obtain a plurality of image slices; carrying out difference filtering and thresholding processing on each image slice; arbitrarily selecting two or more from all the difference images in each image slice, obtaining a plurality of joint probability distribution matrices corresponding to the image slices by using the description of a pixel neighbourhood relationship, and combining elements of each matrix into a one-dimensional vector serving as a feature vector; marking the feature vectors according to category and inputting same into a categorizer to obtain a categorizer model; and extracting, according to the above-mentioned steps, the feature vector of a video to be analyzed and inputting same into the categorizer model to be categorized to obtain a steganalysis result. The present invention takes full advantages of the time domain correlation of a video, so that the steganalysis effect is improved, and the method can be applied to analysis systems of various types of video steganography algorithms.