单曝光HDR图像生成技术研究
发布时间：2019-02-18 14:07
【摘要】：随着显示技术的不断更新,HDR显示设备开始进行推广,对于HDR资源特别是动态的高动态范围(High Dynamic Range,HDR)资源的需求也越来越大,如何高效便捷的获得HDR图像资源正日益成为一个研究的热点,并仍然是一个研究的难点。为了在进行图像捕获时更好地构建HDR图像以及将现有低动态范围(Low Dynamic Range,LDR)图像转换为HDR图像,本文首先提出了一种基于光源采样的单曝光HDR图像生成算法。该算法基于人眼视觉系统(Human Visual System,HVS)模型,分别提取出LDR图像的亮度分量和色度分量。对LDR图像的亮度分量进行反色调映射。然后再对反色调映射后图像的高光区域进行采样,将采样结果作为图像中的光源进行扩展。在扩展过程中为模拟光线散射效应及尽可能地保持该区域的细节,进行高斯滤波和腐蚀操作。最后,在色度分量与亮度分量的图像融合过程中,对图像进行调整,进一步拉伸图像暗区域的对比度,并对暗区域噪声起到一定程度的抑制作用。实验表明,该算法能通过单曝光LDR图像生成HDR图像,处理效果较好,能满足实时性要求。而针对单曝光LDR图像细节信息不足的问题,本文还提出了一种基于细节层分离的单曝光HDR图像生成算法。该算法结合HVS模型,将LDR图像分离出亮度分量和色度分量。对分离出来的亮度分量进行伽马校正,之后再对伽马校正后的图像采用滤波操作,进行细节层分离。然后将分离出来的细节层与亮度分量分别进行反色调映射后进行融合得到新的亮度分量。最后将各图像分量融合得到最终的HDR图像。实验表明该算法能挖掘出部分隐藏的图像细节信息,处理效果较好,能满足实时性要求,具有较好的鲁棒性。
[Abstract]:With the continuous updating of display technology, HDR display equipment has been popularized, and the demand for HDR resources, especially for dynamic high dynamic range (High Dynamic Range,HDR) resources, is also increasing. How to obtain HDR image resource efficiently and conveniently is becoming a hot research topic, and it is still a difficult point. In order to better construct HDR images and convert the existing low dynamic range (Low Dynamic Range,LDR images to HDR images, a single exposure HDR image generation algorithm based on light source sampling is proposed in this paper. Based on the human visual system (Human Visual System,HVS) model, the algorithm extracts the luminance component and chroma component of LDR image respectively. The brightness component of LDR image is inversely mapped. Then, the high-light region of the image is sampled, and the sampling result is extended as the light source in the image. In order to simulate the light scattering effect and keep the detail of the region as much as possible, Gao Si filter and corrosion operation are carried out during the expansion process. Finally, in the process of image fusion of chrominance component and luminance component, the image is adjusted to further stretch the contrast of the dark area of the image, and the noise of the dark area is suppressed to a certain extent. Experiments show that the algorithm can generate HDR image by single exposure LDR image, and the processing effect is good, which can meet the requirement of real time. In order to solve the problem of lack of detail information in single exposure LDR image, a single exposure HDR image generation algorithm based on detail layer separation is proposed in this paper. Combined with HVS model, the algorithm separates the luminance component from the chroma component of the LDR image. Gamma correction is performed on the separated luminance component, and then filtering operation is used to separate the detail layer of the image after gamma correction. Then the separated detail layer and the luminance component are mapped inversely and fused to obtain the new luminance component. Finally, the final HDR image is obtained by fusion of each image component. Experimental results show that the proposed algorithm can mine some hidden image details, have better processing effect, meet the real-time requirements, and have better robustness.
【学位授予单位】：西南科技大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.41
本文编号：2425917
[Abstract]:With the continuous updating of display technology, HDR display equipment has been popularized, and the demand for HDR resources, especially for dynamic high dynamic range (High Dynamic Range,HDR) resources, is also increasing. How to obtain HDR image resource efficiently and conveniently is becoming a hot research topic, and it is still a difficult point. In order to better construct HDR images and convert the existing low dynamic range (Low Dynamic Range,LDR images to HDR images, a single exposure HDR image generation algorithm based on light source sampling is proposed in this paper. Based on the human visual system (Human Visual System,HVS) model, the algorithm extracts the luminance component and chroma component of LDR image respectively. The brightness component of LDR image is inversely mapped. Then, the high-light region of the image is sampled, and the sampling result is extended as the light source in the image. In order to simulate the light scattering effect and keep the detail of the region as much as possible, Gao Si filter and corrosion operation are carried out during the expansion process. Finally, in the process of image fusion of chrominance component and luminance component, the image is adjusted to further stretch the contrast of the dark area of the image, and the noise of the dark area is suppressed to a certain extent. Experiments show that the algorithm can generate HDR image by single exposure LDR image, and the processing effect is good, which can meet the requirement of real time. In order to solve the problem of lack of detail information in single exposure LDR image, a single exposure HDR image generation algorithm based on detail layer separation is proposed in this paper. Combined with HVS model, the algorithm separates the luminance component from the chroma component of the LDR image. Gamma correction is performed on the separated luminance component, and then filtering operation is used to separate the detail layer of the image after gamma correction. Then the separated detail layer and the luminance component are mapped inversely and fused to obtain the new luminance component. Finally, the final HDR image is obtained by fusion of each image component. Experimental results show that the proposed algorithm can mine some hidden image details, have better processing effect, meet the real-time requirements, and have better robustness.
【学位授予单位】：西南科技大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.41
【参考文献】
相关期刊论文 前4条
1 朱恩弘;张红英;吴亚东;霍永青;;单幅图像的高动态范围图像生成方法[J];计算机辅助设计与图形学学报;2016年10期
2 梁云;莫俊彬;;改进拉普拉斯金字塔模型的高动态图像色调映射方法[J];计算机辅助设计与图形学学报;2014年12期
3 霍永青;彭启琮;;高动态范围图像及反色调映射算子[J];系统工程与电子技术;2012年04期
4 刘衡生;沈建冰;;基于亮度分层的快速三边滤波器色调映射算法[J];计算机辅助设计与图形学学报;2011年01期
相关博士学位论文 前1条
1 吴金建;基于人类视觉系统的图像信息感知和图像质量评价[D];西安电子科技大学;2014年
，本文编号：2425917
本文链接：https://www.wllwen.com/kejilunwen/ruanjiangongchenglunwen/2425917.html
