【作者】 武晓斌；

【导师】 赵亦工；

【作者基本信息】 西安电子科技大学 ， 模式识别与智能系统， 2014， 硕士

【摘要】 图像拼接技术是将几幅具有一定重合度的图像拼接成一幅图像。近年来随着该技术在遥感图像、军事、医疗、室内多媒体、娱乐等场所的普遍应用,图像拼接算法也得以快速发展。整个系统基于Altera公司的FPGA芯片实现。在FPGA芯片外挂镁光公司的DDR2 SDRAM外部存储器进行数据缓存。本文对外部进来的图像数据进行了坐标变换、图像拼接等操作,最后将得到的拼接图像通过DVI接口输送到液晶显示器进行显示。坐标变换是指将一幅图像的图像坐标投影到另外一种坐标系中,主要有球面投影,圆柱面投影,立方体投影。由于球面投影和立方体投影的运算量比较大,对于相机的视场角要求比较高,这样会造成高成本。基于上述考虑,柱面投影的算法简单,易于硬件实现,因此本文采用圆柱面投影算法在FPGA上进行实现。图像配准是指对有重合部分的图像通过一定的算法处理,得到两幅图像的平移量,旋转量等形变量。本文中介绍了基于相位的匹配方法,基于灰度特征的匹配方法两种图像配准算法。图像融合是指通过特定的方法将两幅或者多幅具有同一场景,不同特性的图像进行融合成一幅具有多种图像特性的图像。在本文中图像融合主要是为了处理两幅图像的拼接缝处的像素值,使得一幅图像到另外一幅图像平滑过渡,以达到消除拼接缝的目的。基于FPGA实现的图像拼接系统具有实时性好,易于实现,成本较低等特点,这将在安防系统、跟踪系统等方面有很大的应用价值。更多还原

【Abstract】 Image stitching is an approach that splices several images with a certain degree of overlap into a single image. In recent years, as this approach is widely used in remote sensing images, military affairs, medical science, indoor multimedia, entertainment venues, the image stitching algorithms have gotten developed rapidly.The whole system is realized on Altera’s FPGA. Around the FPGA, there are several DDR2 SDRAM chips used for image date cache.In this paper, the image date is processed by several classical approaches, like coordinate transformation, image stitching and then displayed the processed image on LCD.The coordinate transformation is to transform a kind of coordinate to another kind of coordinate. There are three main methods of coordinate transformations: spherical projection, cylindrical projection and cubic projection. However, the spherical projection and cubic projection will cause high cost because the algorithms of them are very complex and they require a large view angle of the camera, which is difficult to achieve. Therefore, in this project, the cylindrical projection is focused on and then it is realized on FPGA.Image registration is mainly about to process images which covered the same areas and then get some data like translation and rotation. In this paper, two algorithms about image registration are introduced. One is based on Phase-Matching Method and the other is based on Gray Feature Matching Method.Image fusion is an approach that aims at fusing several images with the same scene and different characteristics into an image. In this paper, the method of image fusion is used to remove the pixel values on joint together place. And then it realize the purpose that smooth transition from one image to another.FPGA-based image stitching system is easy algorithm, lower cost and more real-timeand this approach described in this paper will have significant applications in security and tracking systems.更多还原