Research On Generating Method Of Motion Video Summary Based On Visual Saliency Detection

With the explosive growth of information on the network,there is also gradually increasing video data.If these large amounts of data information can be efficiently retrieved,browsed,and stored,people can easily obtain the content of interest or the main content of the data.Among visual,tactile,and hearing information,visual information is the most intuitive way for humans to receive information content.Based on this,the summary technology of video information through visual saliency came into being.Through summary processing,you can get information that refine and contains the main of video.In this paper,through the analysis of the video of the motion scene,combined with the video saliency detection to complete the summary generation,the main content can be summarized as:(1)Based on the classical visual attention pyramid model,an improved center-surround pyramid visual model with variable ratio and double contrast calculation is proposed.This method takes the video object saliency detection as the starting point to detect the continuous image sequence obtained from the video.First,calculate the RG,BY,and brightness channel color results of all video frames in the RGB color space,and divide the image into pixel blocks in superpixel mode to speed up the image calculation,and then build a central pyramid through an improved variable pyramid ratio space compared with the surrounding pyramid,two contrast algorithms are used to detect bright or dark objects in different color backgrounds to obtain the difference in image contrast characteristics and merge them.Finally,combined with the extraction results of optical flow motion information,the saliency results of the static image and the dynamic image are combined to obtain the final detection result.Under the condition of following the human visual processing mechanism,this method effectively extracts the salient foreground objects in the image through the new visual attention model,and merges the motion features of the continuous image sequence to further improve the detection effect of the saliency area.(2)A method for extracting key frames of motion video based on multiple image feature extraction and image similarity calculation is proposed.This method selects the conversion of the original video image under the HSV color model and quantizes and fuses the color feature results and image texture results in proportion,and enhances the effect of the previously obtained saliency results.Complete similarity judgment between adjacent images through distance calculation and perceptual hash function.After the cluster filtering,the final video summary is generated.This method can effectively extract the region of interest in the image,and finally get the compressed and refined video summary expressed in the form of key frame image sequence.