Self-adaptive invisible watermark synchronous detection method based on deep learning

Abstract

The invention relates to an adaptive invisible watermark synchronous detection method based on deep learning. The method is based on an Adam optimizer, a frame comprises two U-shaped sub-networks anda CNN sub-network, and the two U-shaped sub-networks and the CNN sub-network are respectively a mark network, a detection network and a discriminator. The marking network takes square areas with fixedsizes at four corners of an extracted watermark image as input of the marking network, outputs four residual images with position information, sequentially puts the residual images back to an original image to obtain a marking image, and sends the shot marking image to the detection network during marking detection, so the detection network detects the watermark image; a mask image for displayingthe position of the residual image is outputted so as to determine four corner points of the watermark image, synchronization of the watermark image and the shot mark image is realized by using perspective transformation, and the original watermark image and the watermark mark image are distinguished by using a discriminator. According to the adaptive invisible watermark synchronous detection method based on deep learning, watermark decoding accuracy is greatly improved on the premise of ensuring the visual effect.

Description

technical field [0001] The invention relates to the technical field of image processing, in particular to an adaptive invisible watermark synchronous detection method based on deep learning. Background technique [0002] At present, the biggest problem faced by most robust image watermarking is geometric attack, and most of the existing watermarking technologies are difficult to resist geometric transformation attacks, such as rotation, scale transformation, etc., because geometric attacks destroy the synchronization of watermark components, even small amplitude Any image rotation or scale transformation may cause the watermark detection process to fail. [0003] Moreover, in the process of taking watermarked images and decoding them, due to the influence of shooting distance, angle and other factors, the taken watermarked images usually have certain distortion and deformation compared with the original image, which is also called geometric deformation. distortion. The geo...

AI Technical Summary

Problems solved by technology

However, the introduction of visible markers will reduce the visual effect of the watermark and increase the risk of the watermark being perceived

Images