摘要：

现有视频图像检索技术大多是基于2幅图片的整个信息比较，计算量大，且未生成有效的视频结构。针对该问题，提出一种基于视频元数据信息的快速图像检索方法。利用交互信息量提取关键帧，将关键帧划分为场景和主题2个部分，计算其灰度分布和概率密度，构造视频元数据，通过元数据信息进行视频检索。实验结果表明，该方法能快速准确检索出用户所需的视频内容。

关键词: 元数据, 关键帧, 交互信息量, 视频检索

Abstract:

Existing video image retrieval technology compares with whole information based on two pieces of picture, has large calculated amount, and it does not generate effective video structure. Aiming at this problem, this paper presents a fast image retrieval based on video metadata information. It extracts key frames from interaction information amount, divides key frame into two parts of scene and theme, calculates their gray-scale distribution and probability density, generates video metadata, and goes on video retrieval through metadata information. Experimental results show that this method can research video content for user quickly and correctly.

Key words: metadata, key frame, mutual information quantity, video retrieval

中图分类号: 

• RP393