基于视觉注意力模型的广告标志插入和变换方法及系统技术方案

一种基于视觉注意力模型的广告标识自动插入和变换方法及系统，其包括：首先，基于所构建的视觉注意力模型预测用户对视频每帧各区域的兴趣区域和对各帧的关注程度；然后，根据用户对各帧的关注程度曲线来确定插入广告的时间点，并通过基于预测出的注意力分布来评价在各区域插入广告的合适程度，进而获取可供广告插入的候选区域序列，并将广告插入到对视频内容影响不大的区域；最后，根据预测出的注意力分布将广告标志插入到合适的时间点和位置，并对其进行多种特征变换，以使其能够重复吸引用户或观众的注意力。根据本发明专利技术可以有效进行广告标识的自动插入和变换，并使所插入的广告标识在不影响正常观看的情况下能够重复吸引人的注意力。

Advertisement mark insertion and transformation method and system based on visual attention model

A kind of automatic insertion of visual attention model based on advertising logo and transformation method and system, which includes: first, visual attention model is constructed to predict the user region of interest in video each frame of each region and the degree of concern for each frame based on; then, to determine the time point into advertising according to user's degree of attention to the curve frame, and based on the prediction of the distribution of attention to the evaluation into the appropriate degree of the advertisement in each region, and then obtain the candidate region sequence for insertion of ads, and advertising into the area with little impact on the video content; finally, according to the prediction of the distribution of attention will sign advertising into the appropriate time and position, and a variety of features of the transformation, so that it can attract the attention of the audience or the user to repeat. According to the invention, the automatic insertion and transformation of the advertisement mark can be effectively carried out, and the inserted advertisement mark can repeat the attractive attention without affecting the normal viewing.

【技术实现步骤摘要】

【技术保护点】
一种基于视觉注意力模型的广告标志插入和变换方法，其特征包括以下步骤：　步骤１根据用户的视觉注意力模型，预测用户对视频每帧不同区域的感兴趣程度以及对各帧的关注程度；　步骤２根据对各帧的关注程度曲线，确定广告标志插入的时间点；　 　步骤３对可以插入广告标志的帧，根据预测出的对每帧各区域的感兴趣程度，选择允许广告标志插入的候选区域序列；　步骤４根据用户的视觉注意力模型，对广告标志进行变换以确保其为最显著区域，并将变换后的广告标志叠加到视频帧上。

【技术特征摘要】

【专利技术属性】
技术研发人员：田永鸿，李甲，黄铁军，高文，
申请(专利权)人：北京大学，
类型：发明
国别省市：11[中国|北京]