图像处理毕业设计论文模版--视频序列图像分割及阴影抑制算法的研究.doc

PAGE 1 安徽建筑工业学院 毕 业 设 计 (论 文) 课 题 视频序列图像分割及阴影抑制 算法的研究 摘 要 在智能视频监控领域、影视技术、多媒体应用技术中，常常需要检测出人体或其它物体，并将其与背景分离，即解决实时背景下目标的分割问题。视频图像的目标分割结果，将对目标分类、跟踪及行为理解等后续处理产生重要影响。 图像分割多年里一直受到研究人员的重视，也提出了数以千计的算法。现今比较流行的目标分割的方法，有不少是忽略阴影检测的，目标总是与阴影一起被检测出来。阴影会引起目标的合并、目标形状的失真等一些严重问题，引起分割和跟踪错误。由于阴影直接影响目标的检测，成为影响后续处理效果的关键因素，有必要进一步研究。 本课题拟根据图像处理的理论基础，对一些传统的边缘检测算子进行了理论分析，用仿真实验测试其边缘检测的效果，对比分析各边缘检测算法效果。介绍几种常用的彩色空间以及彩色空间的转换算法，系统地阐述了图像分割的各种方法，分析总结了几种常用分割方法的优缺点。选用RGB彩色空间，利用背景差分法对图像初步分割后，再利用区域生长法去除目标外部的噪声，分割出带影子的目标图像。然后，分析总结了阴影检测的基本假设和一般框架，及国内外目前主流的阴影检测与抑制算法，指出了这些方法用于去除目标阴影时存在的问题。 针对不同图像的阴影和目标体的特点，拟设计一种去除阴影的算法。基于边缘信息的阴影抑制算法适用于目标体边缘信息丰富，阴影边缘信息相对简单的阴影去除。 关键词 图像分割 阴影抑制 Abstract In the field of intelligent video surveillance, video technology, multimedia technology, often need to detect a human body or other objects, separate them with background, that is the context of solving real-time target segmentation. Video image object segmentation results, will target classification, tracking and behavior understanding such an important impact on subsequent processing. Image segmentation has been for many years in research attention, also raised thousands of algorithms. Goal of the current popular methods of segmentation, shadow detection, many are neglected, the goal is always to be detected, together with the shadow. The merger will cause the shadow of goals, objectives and some distortion of the shape of a serious problem, causing segmentation and tracking error. As the shadow directly affect target detection, a follow-up treatment effect affecting the key factors, the need for further research. The aim of this theory based on image processing based on some of the traditional edge detector is theoretically analyzed, using simulation experiments to test their effect on edge detection, contrast analysis of the effect of edge detection algorithm. Introduce some commonly used color space and color space conversion algorithm, systematically expounded the various methods of image segm