一种基于帧间变化特征的视纹提取、查找比对方法及系统技术方案

本发明专利技术涉及一种基于帧间变化特征的视纹提取、查找比对方法及系统。基于帧间变化特征的视纹提取方法经过有效的视频预处理，尽量将各种不同质量和分辨率的视频统一到一致的分辨率和视频质量水平上。然后通过相邻若干帧之间相关性的计算，找到相邻若干帧之间特征明显的（例如差异最大）的若干图像块，对这些块的编号进行编码，形成帧指纹。对帧指纹序列进行组合，形成视纹。同时，由于本发明专利技术视纹提取方法具有良好的鲁棒性，发明专利技术了与该视纹提取方法配套的，基于倒排查询和窗口滑动匹配的视纹比对查找算法。在保证查询准确度的情况下，极大地提高了检索速度。可适用于实时操作。

Visual line extraction, search and comparison method and system based on inter frame variation feature

The invention relates to a method and a system for extracting, searching and comparing images based on inter frame variation characteristics. The method of visual line extraction based on the inter frame change feature, after the effective video preprocessing, tries to unify the video with different quality and resolution to the same resolution and video quality level. Then, through the calculation of the correlation between adjacent frames, several image blocks with obvious characteristics (for example, the biggest difference) between adjacent frames are found, and the number of these blocks is coded to form frame fingerprints. Frames are combined to form visual lines. At the same time, because of the good robustness of the method of visual line extraction in the invention, an image alignment search algorithm based on inverted query and sliding window matching is developed, which is matched with the visual line extraction method. In the case of ensuring query accuracy, the retrieval speed is greatly improved. It can be used in real-time operation.

【技术实现步骤摘要】

【技术保护点】
一种基于帧间变化特征的视纹提取方法，包括以下步骤：　Ａ．视频预处理：对所处理的视频进行预处理，以降低视频噪声和高频能量；　Ｂ．帧间相关性计算：对视频帧进行分块，计算相邻帧间的相关性；　Ｃ．帧指纹编码：选择相邻若干帧之间差异变化最大的或 差异变化最小的或符合确定规则的若干图像块，对这些块的编号进行编码，形成帧指纹；　Ｄ．视纹生成：对所述帧指纹的序列进行组合，生成视纹。

【技术特征摘要】

【专利技术属性】
技术研发人员：黄铁军，黄铮，田永鸿，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]