多曝光图像融合中的运动检测与鬼影去除方法研究
发布时间：2018-08-01 10:01
【摘要】：伴随着数码摄像技术的不断发展,数字图像越来越多地出现在人们的生产、生活和科学研究当中。近年来,高动态范围成像技术的研究和进展,极大地促进了数字成像技术朝着高清晰度和高信息量方向的发展。同时,这一技术也被越来越多地应用到数字摄像、遥测遥感、安防监控等领域。本文在阐述了数字成像、高动态范围成像、多曝光图像融合、鬼影消除等技术原理的基础上,总结了现有的多曝光图像融合算法,并基于操作域与参考图像对算法进行了分类和比较。在此基础上提出了两种新的多曝光图像融合中的运动检测与鬼影消除算法。本文的研究内容和主要工作如下:1、深入研究了高动态范围成像技术,尤其是多曝光图像融合技术。阐述了动态场景下运动检测与鬼影去除技术的理论基础、国内外研究现状及当前的技术难点,并针对现有的运动检测与鬼影去除算法进行了分类和比较;2、首次提出了一种高效的基于类内一致性与类间一致性的"鬼影"检测和去除算法,以实现高动态范围(HDR)图像的无鬼影重建。通过对选定的参考图像和图像序列中的其他图像进行直方图映射匹配运算,代替以往直接对图像数据操作来检测运动的算法。这种方法可以使场景中的大部分细节得以保留下来,尤其是图像中过曝光或欠曝光的区域,并大大降低了后期运动检测的难度。此外,考虑到不同图像同一位置的像素间的内部一致性与同一幅图像里相邻像素间的外部一致性,将运动检测模型建立在超像素层面。该方法有效地对输入的低动态范围图像序列进行了校准,同时在去除鬼影的前提下最大限度地保留了图像中的细节;3、首次提出一种基于结构一致性与对比度质量的运动检测与鬼影去除算法。通过比较其他图像与参考图像之间的结构一致性大小来对运动区域进行检测。为了将场景中的细节更多的保留下来,引入了对比度指标来进行可见性评价,并在对比度地图的指导下,将中间图像序列进行无缝融合,生成一幅无鬼影的图像,同时大部分场景中的细节得以保留。此外,该方法大大降低了传统算法的运算量。
[Abstract]:With the development of digital camera technology, more and more digital images appear in people's production, life and scientific research. In recent years, the research and development of high dynamic range imaging technology has greatly promoted the development of digital imaging technology towards high definition and high information content. At the same time, this technology is more and more used in digital camera, remote sensing, security monitoring and other fields. Based on the introduction of the principles of digital imaging, high dynamic range imaging, multi-exposure image fusion and ghost elimination, this paper summarizes the existing multi-exposure image fusion algorithms. The algorithm is classified and compared based on the operation field and reference image. On this basis, two new motion detection and ghost cancellation algorithms in multi-exposure image fusion are proposed. The research contents and main work of this paper are as follows: 1. The high dynamic range imaging technology, especially the multi-exposure image fusion technology, is studied in depth. This paper expounds the theoretical basis of motion detection and ghost removal technology in dynamic scene, the current research situation and the current technical difficulties at home and abroad. Based on the classification and comparison of the existing motion detection and ghost removal algorithms, an efficient "ghost shadow" detection and removal algorithm based on intra-class and inter-class consistency is proposed for the first time. In order to achieve a high dynamic range of (HDR) images without ghost reconstruction. By using histogram mapping matching operation on the selected reference image and other images in the image sequence, the motion detection algorithm is replaced by the previous direct operation of the image data. This method can keep most of the details in the scene, especially in the over-exposed or under-exposed areas of the image, and greatly reduce the difficulty of post-motion detection. In addition, considering the internal consistency between pixels in the same location of different images and the external consistency between adjacent pixels in the same image, the motion detection model is established at the super-pixel level. This method can effectively calibrate the input image sequences with low dynamic range. At the same time, under the premise of removing the ghost image, the details of the image are kept to the maximum extent. A motion detection and ghost removal algorithm based on structure consistency and contrast quality is proposed for the first time. The motion region is detected by comparing the structure consistency between other images and reference images. In order to keep more details in the scene, the contrast index is introduced to evaluate the visibility, and under the guidance of contrast map, the intermediate image sequence is seamlessly fused to produce a non-ghost image. At the same time, the details of most of the scenes are preserved. In addition, this method greatly reduces the computational complexity of the traditional algorithm.
【学位授予单位】：山东大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.41
本文编号：2157214
[Abstract]:With the development of digital camera technology, more and more digital images appear in people's production, life and scientific research. In recent years, the research and development of high dynamic range imaging technology has greatly promoted the development of digital imaging technology towards high definition and high information content. At the same time, this technology is more and more used in digital camera, remote sensing, security monitoring and other fields. Based on the introduction of the principles of digital imaging, high dynamic range imaging, multi-exposure image fusion and ghost elimination, this paper summarizes the existing multi-exposure image fusion algorithms. The algorithm is classified and compared based on the operation field and reference image. On this basis, two new motion detection and ghost cancellation algorithms in multi-exposure image fusion are proposed. The research contents and main work of this paper are as follows: 1. The high dynamic range imaging technology, especially the multi-exposure image fusion technology, is studied in depth. This paper expounds the theoretical basis of motion detection and ghost removal technology in dynamic scene, the current research situation and the current technical difficulties at home and abroad. Based on the classification and comparison of the existing motion detection and ghost removal algorithms, an efficient "ghost shadow" detection and removal algorithm based on intra-class and inter-class consistency is proposed for the first time. In order to achieve a high dynamic range of (HDR) images without ghost reconstruction. By using histogram mapping matching operation on the selected reference image and other images in the image sequence, the motion detection algorithm is replaced by the previous direct operation of the image data. This method can keep most of the details in the scene, especially in the over-exposed or under-exposed areas of the image, and greatly reduce the difficulty of post-motion detection. In addition, considering the internal consistency between pixels in the same location of different images and the external consistency between adjacent pixels in the same image, the motion detection model is established at the super-pixel level. This method can effectively calibrate the input image sequences with low dynamic range. At the same time, under the premise of removing the ghost image, the details of the image are kept to the maximum extent. A motion detection and ghost removal algorithm based on structure consistency and contrast quality is proposed for the first time. The motion region is detected by comparing the structure consistency between other images and reference images. In order to keep more details in the scene, the contrast index is introduced to evaluate the visibility, and under the guidance of contrast map, the intermediate image sequence is seamlessly fused to produce a non-ghost image. At the same time, the details of most of the scenes are preserved. In addition, this method greatly reduces the computational complexity of the traditional algorithm.
【学位授予单位】：山东大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.41
【参考文献】
相关博士学位论文 前1条
1 陈浙泊;高动态范围成像技术的研究[D];浙江大学;2009年
相关硕士学位论文 前2条
1 郭剑桥;多曝光融合算法研究及HDR图像合成与色调映射算法设计实现[D];中国科学技术大学;2016年
2 朱良销;高动态范围成像系统的研究[D];电子科技大学;2011年
，本文编号：2157214
本文链接：https://www.wllwen.com/shoufeilunwen/xixikjs/2157214.html
