面向颜色空间的彩色图像边缘提取研究.doc
《面向颜色空间的彩色图像边缘提取研究.doc》由会员分享,可在线阅读,更多相关《面向颜色空间的彩色图像边缘提取研究.doc(11页珍藏版)》请在三一文库上搜索。
1、面向颜色空间的彩色图像边缘提取研究 摘 要 人们在获取外界知识时,视觉系统所能提供的信息不仅直接,而且作用更为明显,大约能够占到所有信息量的80%。在人们的生产和实践中,人类可以通过眼睛获取的视觉信息完成很多任务,而大部分信息的传递都要依靠图像的参与。 现实世界是丰富多彩的,图像中不同的目标或区域之间,信息也是不连续的,颜色有或多或少的过度。图像也通过颜色的不同来勾勒物体的基本轮廓,这就涉及到了图像中的边缘。边缘是图像的重要特征之一,是图像局部亮度变化中最显著部分。我们可以通过对边缘的研究,处理不同需求或不同类型的图像。图像处理技术有很多种,对于那些颜色信息类似的边缘像素点的处理,在人们丰富多
2、彩的生活中有着广泛的应用,因此人们也非常的重视。例如,车牌定位中能把车辆冲公路背景中成功的分离出来,实现目标车辆的追踪检测;农业中也可以用来检测户外植物的生长;另外对合成孔雷达图像等特殊图像的研究也越来越重视,学者们对图像边缘检测评价的研究日益关注。因此,彩色图像边缘的详细检测和正确的定位就显得尤为重要。 本文先介绍了早期人们对于灰度图像边缘检测的算法,并对经典的检测算子进行了分析比较,在后边给出了适当参数下对lena图像的检测结果;然后简单列举了几个常用的颜色空间,并对介绍了已有的部分彩色图像的边缘检测方法,如基于某空间对经典算子的扩展,小波多尺度下真彩色图像的检测,对结合神经网络的算法也进
3、行了分类研究。在这些基础上,本文的具体工作如下: 1、提出了一种新的基于颜色三角形利用面积把矢量换成标量的彩色图像边缘检测算法。在这个算法中,我们把像素点颜色信息的变化转为三角形的面积变化,把满足条件的彩色信息转化为三角形的三边关系,利用边缘检测算子实现对面积的界定,再通过阈值的比较确定边缘信息。既全面的考虑了彩色分量之间的联系,还把抽象的向量交互转化到对标量的计算,有较强的理论意义,实验结果表明边缘定位比较准确而清晰。 2、作者结合细胞神经网络,将三角形的颜色信息作为输入,在确定图像边界以后,通过卷积,判断网络是否完全收敛,进而定位边缘。实验结果证明,这种方法得到的边缘信息丰富且精确,同时,
4、消除了噪声。 3、在经过对三角形信息量的分析之后,提出了一种基于直角梯形信息量的算法,弥补了图像中部分像素点颜色分量不能构成三角形,造成部分边缘丢失或错误的缺陷。实验结果证明,与基于三角形信息量的算法相比,它消除了部分错乱边缘,结果更加清晰。 关键词:彩色图像边缘检测;RGB颜色空间;Sobel算子;颜色三角形;颜色梯形;CNN Abstract When people get external knowledge, vision systems can not only provide information directly, but also effect more apparently
5、, which can account for about 80% of all information. In the peoples production and practice, human can complete a lot of tasks through visual information that obtained by their eyes, while most of the information transfer must rely on the participation of the image. The real world is rich and color
6、ful, between different targets or regions in an image, the information is also discontinuous, and colors are more or less excessive. The images also draw the outline of the basic objects through different colors, which relates to the edge of the image. Edge is one of the important features of the im
7、age, which is the most significant part of local brightness changes in image. Through the research on edge, we can deal with different types of images or there different requirements. There are varieties of image processing technologies, and for the processing of those edge pixels that have similar
8、color information, it has a wide range of applications in peoples colorful life, therefore people are very serious. For example, the vehicle license plate location can successfully separate vehicles from highway backgrounds to achieve target vehicle tracking detection; In agriculture, it also can be
9、 used to detect the growth of outdoor plants; In addition, people also pay more and more attention to the research of synthetic aperture radar image and other special images, and also the increasing concern of scholars on the evaluation of image edge detection. Therefore, detailed color image edge d
10、etection and correct positioning is particularly important. This paper first introduces peoples research on early gray-scale image edge detection algorithm, analyzes and compares the classical detection operator, and presents the detection results of lena image with appropriate parameters; Then simp
11、ly lists some commonly used color space, and introduces the existing part of the color image edge detection method, such as expansions to classical operators based on a certain color space, wavelet multi-scale detection of true color images, algorithms combined with neural network are also researche
12、d. Based on all the above, the concrete works are as follows: 1、It proposes a new color image edge detection method using color triangle, in which vector is replaced by scalar using area. In this algorithm, change of color informations in the pixels are replaced by change of area of the triangles, a
- 配套讲稿:
如PPT文件的首页显示word图标,表示该PPT已包含配套word讲稿。双击word图标可打开word文档。
- 特殊限制:
部分文档作品中含有的国旗、国徽等图片,仅作为作品整体效果示例展示,禁止商用。设计者仅对作品中独创性部分享有著作权。
- 关 键 词:
- 面向 颜色 空间 彩色 图像 边缘 提取 研究
链接地址:https://www.31doc.com/p-9186543.html