Research On Image And Video Retargeting Based On Adaptive Blocking

With the continuous emergence of various digital products,product terminals have different screen display ratios,which may cause content distorted,lost,or incapable of fullscreen,seriously affecting the visual experience,when the same picture or the same video display on different devices.The content-aware image retargeting method can limit the distortion or loss of image or video content to the region which is not easy to be noticed by human eyes while retargeting,so that the retargeted visual effect is obviously superior to the traditional image scaling method.However,the method still has the problem of losing the important content when retargeting the image or video which important area is large,insufficient edge protection for the image or video,and the problem of playing jitter after video retargeting.In view of the above problems,this thesis proposes an image and video retargeting technique based on adaptive blocking.The main work of the thesis includes:In the aspect of image retargeting,the adaptive blocking is integrated into the seam carving(SC),and the cumulative energy map is optimized,which can protect the important area in a certain extent and also take the retargeting effect of the background region into account.First,the original image is divided into protected and unprotected regions according to the average cumulative energy vector of the saliency map.Then the number of seams is allocated according to the area of each region.The SC method has the problem that the minimum energy seam maybe not the optimal seam when a seam is generated,and the seam starting point position is not suitable and needs to be improved.This thesis improves these defects and optimizes the cumulative energy map,which reduces the probability that small areas of saliency objects to be cut off.In the aspect of video retargeting,considering the real time of video retargeting method,this thesis combines the idea of adaptive blocking with the scaling method.The protected and unprotected regions of video frames are divided by adaptively blocking,and then the image is retargeted quickly by scaling algorithm.Moreover,the frame-by-frame optimization based on the scaling ratio is carried out at the same time.According to the Struck tracking algorithm,the video object position of each frame is determined by original image and retargeted image,and the frame-by-frame optimization is carried out according to the position.The problem of video jitter can effectively be controlled.To validate the proposed image retargeting algorithm,the retargeting results are compared with the state-of-the-art SC and its improved algorithms in the MSRA database,and the online subjective evaluation scaling results are performed on the Internet.The evaluation results show that the proposed method has a better subjective retargeting effect for various types of image.The video retargeting algorithm is verified on Visal and SegtrackV2 video databases,and the retargeting results are compared with SC algorithm,and subjective evaluation is performed.The experimental results show that the proposed video retargeting method can improve the temporal correlation of video,and the saliency part of the video subject is protected.