CASIA OpenIR  > 毕业生  > 硕士学位论文
气动光学效应图像校正与复原算法研究
其他题名Research of Image Correction and Restoration Algorithms in Aero-Optic Effect
杨波
学位类型工学硕士
导师张文生
2013-05-24
学位授予单位中国科学院大学
学位授予地点中国科学院自动化研究所
学位专业控制理论与控制工程
关键词气动光学效应 图像校正 图像复原 图像配准 图像融合 Aero-optic Effects Image Correction Image Distortion Image Registration Image Fusion
摘要高速飞行器在大气层中飞行时,光学头罩与高速气流发生剧烈的相互作用,导致其探测窗口附近形成复杂的流场和温度场,引起气动光学效应,使得红外和可见光图像出现模糊与畸变。为了获取目标的原始图像,实现准确的目标识别,必须对模糊与畸变图像进行校正和复原。相关技术在航天、航空、国防等领域有广泛应用,具有重要理论意义和应用价值。针对气动光学效应下的图像模糊、抖动和畸变问题,本学位论文研究了图像配准、图像融合与图像复原理论与方法。 由于气动光学流场和温度场的复杂性,CCD获取的每一幅图像会产生不同程度的模糊与畸变,单幅图像很难有效地去除气动光学效应的影响,获取高保真的原始图像。本文选择了序列多帧图像作为信息源,利用彼此的互补信息进行图像校正和复原,设计了多组实验验证,实验表明了本文算法的有效性。 本文主要工作与贡献如下: (1)针对序列图像的畸变问题,提出了一种基于B样条和RC相似度测量的图像配准算法,以及新的边缘修正算法。考虑到图像像素之间的关联特性,用RC距离代替传统的SSD、 MI距离,在这种新的距离函数上,利用基于B样条的图像配准算法,对序列多幅图像中的每帧图像进行配准;设计了一种新的边缘修正算法,利用该算法消除配准后出现的锐利边缘。选择SIPI-misc数据库进行了4组实验,实验表明:引入新的RC图像距离,用基于B样条的配准算法和新的边缘修正算法,能显著地对序列图像的畸变进行校正。 (2)针对单帧图像校正后存在细节信息不足问题,提出了一种新的基于质量指数的图像融合算法。通过亮度、相关性、对比度等因子,对配准后的图像进行质量评价,基于该质量评价函数,对序列图像进行融合,实现了从序列图像中复原出一幅具有丰富细节的图像。选择上述SIPI-misc数据库的校正图像集合,进行了融合实验,实验表明:该融合算法能有效地从序列图像中恢复图像的细节信息。 (3)对气动光学效应图像的模糊、抖动与畸变过程,进行整体建模,融合图像配准、图像融合和图像复原技术,获取原始图像。在复杂的气动光学流场中,整体考虑CCD获取的图像模糊、畸变、噪声影响,构建了一种整体的退化模型;根据该退化模型,融合上述图像配准和图像融合技术,引进MEPLL(maximizing the Expected Patch Log Likelihood)图像复原算法,对图像的模糊、抖动与畸变影响进行去除。通过与国际权威算法对比,实验结果表明:本文算法在图像的清晰度和形变复原上都有显著提升。
其他摘要During the aircraft flying in the atmosphere with high speed, drastic friction occurs between the aircraft’s optical hood and the high-speed airflow. Complex flow field and temperature field are formed in the vicinity of detection window, producing aero-optical effects. As a result, the images received by the camera on the aircraft are blurring, distorted, or shifting. In order to get the original images and recognize targets precisely, we must correct the distortion and restore the images. Related technologies play an important role in the field of aerospace, aviation, national defense and in many civilian systems, which has important theoretical significance and application prospects. To resolve image blurring, shift and distortion problems happening under aero-optics, this paper does research on distortion correction, image fusion, and image restoration. Due to the complexity of the pneumatic optical flow field and temperature field, complicated distortion and blurring phenomenon happens to the each image obtained by CCD. A single image can’t effectively remove the influence of environmental factors. This paper is based on sequence images, making use of each other’s complementary information to restore a single image. The main work and contributions are as follows: (a) An image registration algorithm based on B-spline and RC image distance has been proposed. Taking into account the non-independence between adjacent pixels, we use RC image distance to substitute traditional image distance such as SSD, MI, based on this new image distance function, use .B-spline based image registration to correct the sequence images. Also, an edge-refining method has been proposed, to deal with sharp edge problem after registration process. By performing 4 group experiments on the SIPI-misc database, it shows that the registration algorithm proposed by this paper can effectively correct the distorted sequence images. (b) An image fusion algorithm based on image quality index has been proposed. Image quality index evaluates image quality from three factors: loss of correlation, luminance distortion, and contrast distortion. Based on image quality index matrix, the fusion process fuses the sequence images to produce an image of rich details Experiments show that this fusion algorithm can recover an image with rich information from degraded sequence images. (c) An overall implementation of restoring image from aero-optics is provided. By modeling the distortion and b...
馆藏号XWLW1911
其他标识符201028014629083
语种中文
文献类型学位论文
条目标识符http://ir.ia.ac.cn/handle/173211/7662
专题毕业生_硕士学位论文
推荐引用方式
GB/T 7714
杨波. 气动光学效应图像校正与复原算法研究[D]. 中国科学院自动化研究所. 中国科学院大学,2013.
条目包含的文件
文件名称/大小 文献类型 版本类型 开放类型 使用许可
CASIA_20102801462908(1535KB) 暂不开放CC BY-NC-SA请求全文
个性服务
推荐该条目
保存到收藏夹
查看访问统计
导出为Endnote文件
谷歌学术
谷歌学术中相似的文章
[杨波]的文章
百度学术
百度学术中相似的文章
[杨波]的文章
必应学术
必应学术中相似的文章
[杨波]的文章
相关权益政策
暂无数据
收藏/分享
所有评论 (0)
暂无评论
 

除非特别说明,本系统中所有内容都受版权保护,并保留所有权利。