|
In this paper, the application of machine learning in video watermarking , support vector machines and ensemble learning application , first classify video watermarking attacks often suffer , and then under each category are given due prevention methods , the focus the the video watermark most vulnerable , but also most of the attacks , collusion attack . Against collusion attack study at this stage at home and abroad were summarized , and the typical algorithm , provides a new direction for the future of collusion attack ideas . According to the video is composed of multiple frames , the different frames may be embedded in the characteristics of the watermark according to different algorithms , and design of the multi watermark Multiple algorithms , in order to improve the robustness of the single algorithm weak defect (i.e., a single algorithm can not resist all attack ) , and video watermark robustness . With clear classification method , and each attack prevention strategy , multiple watermarking algorithm is designed to analyze the prospects for integrated learning in the digital video watermarking meeting point , making intelligent design selected kinds of watermarking algorithm becomes possible . Then , on this basis , according to the characteristics of the video , the video texture characteristics proposed a video watermarking algorithm based on HVS , so as not to reach the best compromise between visibility and robustness , support vector machines in digital watermarking the application laid a foundation. The support vector machine is an effective method of machine learning , primarily used for classification and regression problems , but can also be used . Fit , based on support vector machines and digital watermarking there have been applied to images , audio , video also too little support vector machine , this paper, based on the idea of embedded watermark image applied to video provide new ideas combined with support vector machine in the video watermark embedding watermark .
|