基于视音频特征的体育视频关键场景自动分割方法  

Automatic Segmentation of Key Scenes in Sports Video Based on Video and Audio Features

作　　者：杨丰旭[1] YANG Feng-xu(Heilongjiang University of Chinese Medicine,Harbin 150040 China)

机构地区：[1]黑龙江中医药大学,黑龙江哈尔滨150040

出　　处：《自动化技术与应用》2023年第8期70-73,共4页Techniques of Automation and Applications

摘　　要：近年来,人们对于体育赛事视频的观看需求有所提升,大部分观众希望可以直接观看至赛点内容。传统体育视频关键场景自动分割方法提取特征效果不佳,造成关键场景分割结果准确率降低。针对此问题,设计基于视音频特征的体育视频关键场景自动分割方法。采用傅里叶变换技术以及色彩空间模型获取体育视音频特征,构建噪声估算模型,应用梯度倒数加权法完成视频图像增强处理。使用光流法完成关键场景分割,使用信息熵计算关键场景分割结果与体育视频的一致性。实验结果可得,使用视音频特征后,关键场景分割准确度得到提升。In recent years,people's demand for watching videos of sports events has increased,with most viewers wanting to be able to watch the content right up to the match point.The traditional automatic segmentation method of key scenes in sports video is not effective in extracting features,which leads to the decrease of the accuracy of segmentation results.To solve this problem,an automatic segmentation method for key scenes of sports video based on audio and video features is designed.Fourier transform technology and color space model are used to obtain sports audio and video features,noise estimation model is constructed,and gradient reciprocal weighted method is applied to complete video image enhancement processing.Optical flow method is used to complete the segmentation of key scenes,and information entropy is used to calculate the consistency between the segmentation results of key scenes and sports video.The experimental results show that the segmentation accuracy of key scenes is improved by using audio and video features.

关 键 词：视音频特征 视频分割 图像增强 图像降噪 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]