论文总字数：19217字

摘 要

随着多媒体技术和网络技术的迅速发展，并且多媒体信息的日益剧增，但是如何有效的对这些复杂视频数据进行有效的组织，表达，存储和管理，并且如何对其快速的检索与浏览已经成为视频关键帧提取领域中的一项重要研究内容。由于视频内容的丰富性与其多样性，以及复杂的时空多维结构特性，传统的思维管理方式与检索方案已经不能够满足我们从巨大而繁琐的数据中提取我们所需要的内容，所以便提出了各式的关键帧提取的算法。

1. 通过对原始的视频数据的分析与检测，建立一种简单树形结构模型，该模型能够描绘从代表镜头，镜头聚类，镜头，关键帧，帧聚类到帧的构造化层次，从而构成后续进行视频处理的基础。
2. 通过总结与归纳提出一种新的关键帧提取方案。主要是将亮度直方图与阈值结合起来，该方法概括了突变镜头的检测与渐变镜头的渐变过程，这两个处理过程都是进行相邻两帧的亮度直方图帧差与阈值的比较来自动进行挑选所需要的内容。实验结果表明，改进算法具有很好的检测效果，准确率比较高，且计算复杂度低，易于实现。
3. 针对提出的算法，运用MATLAB软件编出相应的程序，进行实验分析，通过对具体的视频处理来表现提出算法的优越性以及日后需要改进的地方。

关键词：视频检索 镜头检测 关键帧

Research on video key frame extraction algorithm

Abstract

With the rapid development of multimedia technology and network technology, and multimedia information is increasing, but how effective these complex video data effectively organize, express, store and management, and how to rapid retrieval and browsing of video key frame extraction has become one of the important research contents in the field of. Because of the richness and diversity of video content, as well as the complex temporal multidimensional structure characteristics, the traditional thinking mode of management and retrieval scheme has not been able to meet the us from the huge and complicated data to extract the contents, so they put forward the key frames to all kinds of extraction algorithm.

1. Through the analysis and detection on video the original data, establish a simple tree structure model, the model can describe the delegates from the shot, shot clustering, shot, key frame, hierarchical structure frame clustering to frame, which constitute the basis of subsequent video processing.

2. An improved method is proposed by summarizing, the luminance histogram and threshold combined, the method generalizes the process of gradual change detection process abrupt and gradual shot, the two processes are the key frame by comparing the luminance histogram of frame difference of two adjacent frames and threshold to automatically select the need. The experimental results show that, the improved algorithm has very good detection effect, high accuracy, and low computation complexity, easy realization.

3. Using the algorithm proposed in this paper, using MATLAB software to compile the program, experimental analysis, through to the video processing specific to the performance superiority of the proposed algorithm and the future areas in need of improvement.

Key Words: Video retrieval Shot detection Key frame

目 录

摘 要 2

Abstract 3

第一章 绪论 5

1.1 关键帧定义 5

1.2 研究视频关键帧提取技术的意义 5

1.3 关键帧提取研究的基本方法 6

1.4 视频关键帧提取的国内外研究成果 7

1.5 研究问题的手段 8

第二章 视频检索概述 10

2.1 数字视频的基础 10

2.2 视频检索框架 10

第三章 镜头检测 12

3.1 镜头变换的基本概念 12

3.2 传统的镜头分割方法 13

3.3 改进的关键帧提取方法 14

3.3.1 阈值的选取 16

3.3.2 对于变化镜头的检测 16

3.4 模型选择器 18

第四章 系统的编程设计 19

4.1 系统的总体构架 19

4.2 主要程序 19

第五章 实验结果分析与讨论 22

5.1 验证计算公式以及实验结果分析 22

5.2 对于算法的测试 24

5.3 总结 24

参考文献 26

附录一 28

请支付后下载全文，论文总字数：19217字