Video digital watermarking method based on characteristic scale variation invariant points and microscene detection

Abstract

The invention provides a video digital watermarking method based on characteristic scale variation invariant points and microscene detection. The method includes the steps that the characteristic scale variation invariant points in a video frame image are extracted, and it is guaranteed that reference points selected for watermark embedding are evenly distributed in the video frame image through a minimum spanning tree clustering algorithm. Through an algorithm, an annular neighborhood of the characteristic points is used as a watermark embedding region. The invention further provides a watermark embedding mechanism based on scene detection. The content of embedded watermarks varies along with variation of scenes, and good robustness is achieved especially for a video digital watermark attack mode. In experimental measurement, conventional image processing attacks, geometric transformation attacks, frame averaging, frame interposition, frame extraction and the like are conducted on watermark videos respectively. Results show that the method has good robustness and can well resist various common attacks.

Description

technical field [0001] The invention belongs to the field of information processing, in particular to a video digital watermarking method based on video content and involving video image feature points and micro-scene detection. Background technique [0002] In essence, digital watermarking is a communication process, that is, to transmit information from the watermark embedder to the receiver. Therefore, people can learn from the communication transmission model to study digital watermarking. Smith and Comiskey first proposed that the entire watermarking process can be expressed as a basic communication system model. Subsequently, the model has been further developed and has become a standard model for researching digital watermarking systems. figure 1 It is a digital watermarking system model. [0003] The current video digital watermarking technology can be divided into the following three categories according to the method of watermark embedding: [0004] 1. Embed wa...

AI Technical Summary

Problems solved by technology

[0007] Disadvantages: It will increase the data bit rate of the video stream and affect the invariance of the video rate; for the compressed video, it needs to be decoded first, and then re-encoded after embedding the watermark, which increases the complexity of calculation and reduces the quality of the video

[0023] Disadvantages: Due to the limitation of the compression bit rate, the size of the embedded watermark data is limited, and the strength of the embedded watermark is constrained by the video decoding error, and the effect after embedding may change perceivably

[0029] Disadvantages: The encoder and decoder need to be modified, and there is GOP (Group of Picture) error accumulation, which will reduce the quality of the video

[0031] (1) Since the image digital watermarking algorithm has been developed very maturely, there are many kinds of attack methods for image digital watermarking technology, and they are very effective. However, the current video digital watermarking technology does not have a digital watermark that can resist multiple attacks Algorithms, especially geometric transformation attacks

[0032] (2) Most of the video digital watermarking technology is only the expansion and extension of the image digital watermarking technology, and does not fully consider the essential difference between video and image, so it is less robust to the attack methods especially for video

Images