Research On Kev Technologies Of Visible Video Watermark Detection And Removal

With the continuous development of information technology,adding visible watermark has become an effective means to protect the copyright of video in online video media platform.At the same time,the detection and removal of video visible watermark has been widely concerned.Nowadays,the position and shape of video visible watermark on the network will change with time.In this case,using the existing visible watermark detection and extraction methods to obtain the information of its position and shape needs some human intervention,or there are some problems such as incomplete extraction or no watermark information.At the same time,in the existing visible watermark removal methods,the video after watermark removal has the problems of unreasonable edge and wrong filling.In view of the existing problems in the detection and removal methods of video visible watermark,this paper proposes a detection and removal method of video visible watermark based on the algorithm of target detection,image segmentation and image inpainting.The main contents of this paper are as follows:(1)In order to solve the problem of human intervention and difficult to extract dynamic watermark in the process of detection and extraction of video visible watermark,a method of detection and extraction of video visible watermark based on target detection and image segmentation algorithm is proposed.Firstly,the video is obtained from the online video website,and the video visible watermark dataset is constructed.The coordinate position of the watermark in the video frame is detected by the target detection algorithm.According to the obtained coordinate position,a binary mask is generated to mark the watermark position.The watermark part is segmented by image segmentation algorithm to get the watermark pattern.If the overlap degree of the watermark pattern and the real pattern obtained in the current frame is less than a certain value,the frame with the best effect in the previous frames will be processed or operated with the current frame,and the result obtained will be taken as the extraction result of the current frame.The watermark pattern obtained by the watermark extraction part is used for the subsequent image visible watermark removal and video visible watermark removal.Finally,the experimental results of watermark detection and extraction are compared with other classical algorithms.(2)In order to solve the problems of edge irrationality and wrong filling after the removal of visible watermark,an image visible watermark removal method based on the prior of deep image is proposed.Firstly,the extracted watermark is mask preprocessed to get the mask,then the mask with the original video frame are input into the deep image prior network to remove the watermark.In order to solve the problem that the structure of the image generated by the deep image prior network is not clear enough,the structure similarity is introduced into the loss function and the pixel value outside the mask area of the original image is retained.Then on the dataset constructed in this paper,the image visible watermark removal method proposed in this paper is compared with the other three common image inpainting algorithms in subjective and objective aspects respectively.Finally,on the dataset constructed in this paper and LVW dataset,the comparison of this method and the other three algorithms are conducted.(3)In view of the problems of artifacts and unclear texture in the method of image visible watermark removal,a video visible watermark removal method based on optical flow guidance is proposed.Firstly,we use the flow extraction algorithm to extract the optical flow information of the video,detect and extract the watermark pattern of the video,then transform the watermark pattern into a mask,input to the Deep Flow Completion Network together with the optical flow to repair the flow.Finally,under the guidance of the repaired flow information,the video frames are propagated pixel by pixel,and the image watermark removal method proposed in this paper is used to repair the video frames whose pixel propagation is not completely repaired.The experimental results show that compared with the other two algorithms,the PSNR,SSIM and ms-ssim values of the continuous video frames after watermark removal are higher,the local motion information after watermark removal is more reasonable,and the problem of artifacts and unclear structure texture is significantly improved.