【作者】 曹辉；

【导师】 赵丽红；

【作者基本信息】 东北大学 ， 生物医学工程， 2009， 硕士

【摘要】 随着计算机和数字图像处理技术的迅速发展,图像拼接已经成为基于图像绘制技术的一项关键技术,目前,这项技术已经广泛地应用在医学图像处理中。本文侧重研究如何将生物细胞显微图像拼接合成为一幅完整的全景图像。本文在总结现有各种经典图像拼接算法的基础上,根据生物细胞显微图像的图像特征和应用需求,分别针对灰度和彩色两类图像,实现了生物细胞显微图像的准确、快速拼接。论文首先针对灰度图像拼接的应用需求,研究了三种图像匹配算法,第一种方法实现了基于灰度相关的图像匹配算法,针对生物细胞纹理丰富的特点,该算法中提出了多模板匹配方案,提高了生物细胞图像的匹配精度；第二种方法为了提高图像的匹配速度,实现了基于特征的图像匹配算法；第三种方法采用对图像进行二值化处理的算法实现了生物细胞图像的拼接,提高了图像匹配的可靠性和实时性。对以上三种匹配算法,在实时性、精确性两方面进行了分析。在灰度图像平滑算法研究上,采用了利用最小二乘方法建立亮度变换函数,实现两张连续图像的亮度调整和平滑,消除由于重叠图像亮度差异而引起的拼接缝隙,实现图像无缝拼接。其次研究了生物细胞彩色图像拼接算法。针对彩色图像拼接的应用需求,提出了一种基于颜色的色度和饱和度特征来实现生物细胞彩色显微图像拼接的算法,同时提出了一种用于彩色图像匹配中的特征模板提取方法。通过待拼接的两幅彩色图像之间的重叠区域亮度关系分析,建立了图像之间的亮度变化函数,全局化地调整两幅图像之间的颜色差异,消除拼接缝。本文基于现有的研究工作,分析和指出了图像拼接技术的重点发展方向。更多还原

【Abstract】 With the rapid development of computer and Digital Image Processing technology, image mosaic has become the key technology based on image plotting. Currently, the technology has been widely adopted in medical image processing. This paper emphasis on how to make microscopic image of biological cell be a complete panoramic image by image mosaic technology.According to specification and application requirements of microscopic image of biological cell based on current classic image mosaic algorithm, this paper presents improved mosaic algorithm design which focus on grey and color images, on account of this improvement, fast and precisely matching of microscopic images of biological cell is achieved. Firstly, algorithm of image mosaic for grey images of biological cell is talked. Three algorithms of image matching are presented according to application requirement of grey image matching. First of all, a matching algorithms based on gray scale correlativity is implemented. We put forward an algorithm which can elect multi-template according to the different positions of characteristic object and enhance the precision of image registration. Second of all, in order to enhance speed in image registration, we realize the matching algorithm based on feature. Finally, the matching algorithm based on binary is adopted to enhance speed and reliability in image registration. The real-time and precision have been analyzed to the above three algorithms.Furthermore, Normalized Correlation method (NC) is adopted to validate the optimal matching position, by which two consecutive images are mosaicked to an image. An intensity transformation function between the two consecutive images is built to smooth the mosaicked image. Secondly, the paper presents algorithm of image mosaic for color images which using gradated in-and-out blending algorithm to eliminate mosaic gap owing to a difference in brightness and make the image link together seamlessly. In this paper, the key points of development about the technology of image mosaic are analyzed and put forward based on current research work.更多还原