图像拼接技术研究与应用_祝传贝(2)

近年来图像拼接技术的研究与应用随着人们生活以及科学研究的需要正受到越来越多的关注。图像拼接(Image Mosaics)主要是利用现有的计算机等设备,将两幅或者多幅具有相关重叠区域的图像拼成一幅大型的无缝高分辨率图像的技术。本文通过介绍图像拼接技术的图像获取、图像预处理、图像配准以及图像融合等步骤,并利用Visual C++中的MFC编程实现了两幅简单的具有一定重叠区域的位图的拼接。

Abstract:Segmentation of the CT images is a meaningful step in diagnose of clinical injuries and treatment. Snake model can get a better result among varieties of segmentation methods. An optimized segmentation method based on snake model (or parametric active contour model) is proposed. Firstly, an edge map is generated by morphology operation from the original CT images. Then, the proposed function is used to iterate in the segmentation process. The model addresses the problems of unable to converge to concavity and noise sensitivity of the traditional snake models. Andthe model can be applied to practical usage with accurate results.

In recent years, the research and application of image stitching technology along with the people living and the need of scientific research has been paid more and more attention. Image mosaic (Image Mosaics) is mainly the use of existing computer equipment, two or more images together with relevant overlap region a large seamless high resolution image technology. It breaks through the limitation of

traditional image acquisition devices, are widely used in aerospace, medical image analysis, computer vision, video surveillance and other fields, is a very important branch of the current image processing. This paper introduces the image stitching technology through the image

acquisition, image preprocessing, image registration and image fusion step, and realized two pieces of a simple with some overlap region of the bitmap using Visual C++ splicing in MFC programming.

Keywords: CT images, clinical diagnose, image segmentation, morphology, snake model 1 引言

图像拼接技术是将数张有重叠部分的图像拼接成一幅大型无缝的并具有高分辨率的图像。其中，图像配准和图像融合是图像拼接的两个关键技术。对于拼接后的图像要求最大程度的与原始的图像接近，失真尽可能的小，没有明显的缝合线。当人类获得同一场景的两幅或多幅图像时，为了能够获得该场景更多的信息，就必须对图像进行相应的处理。图像拼接是图像处理领域的一个重要分支，图像拼接技术在生产及生活的很多领域中都有广泛的应用。随着研究的不断深入，图像拼接的方法已经有很多了，各种拼接方法都有其自身的特点及相应的应用领域。在2003年，M.Brown发表的Recognising Panoramas文章中提出了基于尺度不变特征技术的SIFT算法，其通过提取关键点来实现图像的配准，该算法完全自动而且拼接的效果好，也是目前国内外研究比较热的算法。我国图像拼接技术起步的比较晚，但是也得到了迅猛发展。其中2004年，赵向阳提出了一种基于Harris角检测算子全自动稳健的图像拼接融合算法，使提取的进度达到了亚像素级；2005年，侯舒维，郭宝龙针在现有的图像拼接技术的基础之上，提出了一种图像拼接技术的快速算法，该算法综合考虑了图像拼接技术的精度和速度。之后，又有很多学者专家在图像拼接领域进行了大量研究，并分别提出了一些改进方法，使图像拼接技术得到了空前的发展。