基于经验小波变换的多曝光图像融合方法的研究

发布时间：2018-05-17 11:00

本文选题：经验小波变换 + 高动态范围成像　；参考：《吉林大学》2017年硕士论文

【摘要】：真实场景以及人眼可识别的亮度范围远远超出普通显示设备的亮度范围。在摄影时,如果真实场景中的亮度范围较大,普通拍摄设备也无法直接捕捉场景中所有的细节信息。高动态范围成像(HDRI,High Dynamic Range Imaging)方法主要分为基于色调映射的方法和基于图像融合的方法。相较于基于色调映射的方法,基于图像融合的方法的计算效率更高,且不需要额外的曝光参数。自从图像融合方法被用来解决高动态范围成像问题并取得令人满意的效果以来,多曝光图像融合问题受到了研究人员越来越的重视。多曝光图像融合方法对HDR场景使用不同曝光度参数进行多次拍摄,通过融合包含不同动态范围细节的图像,尽可能地保留场景中不同动态范围的细节信息,实现高动态范围成像。近年来,不断有新的图像融合方法被使用在高动态范围成像领域。经验小波变换(EWT,Empirical Wavelet Transform)是近年被提出的信号分析和处理方法,它是一种包含经验信息的自适应的小波变换(WT,Wavelet Transform)。经验小波变换先分析信号的傅里叶频谱,然后通过特定的方法生成一组经验滤波器,最后使用生成的经验滤波器组对信号进行滤波。实验表明,经验小波变换可以提取信号的内在模式。二维经验小波变换(2D EWT)方法是经验小波变换的二维版本,可以用来处理图像数据。二维经验小波变换有很多不同的种类,本文主要使用其中最简单的二维张量经验小波变换。本文主要探讨如何将经验小波变换用于处理多曝光图像融合问题,并实现了一种基于经验小波变换的多曝光图像融合方法。RGB图像亮度信息跟R、G和B三个颜色通道都有关系,而亮度又是多曝光图像融合中的重要参考,如果分别对RGB图像的三个通道进行融合,在一些情况下会导致融合结果中出现颜色失真现象。针对这一问题,本文提出一种在L*a*b颜色空间进行的基于经验小波变换的多曝光图像融合方法对上述算法进行改进。由于L*a*b颜色空间中只有L*颜色通道包含亮度信息,算法避免了亮度信息在不同颜色通道上产生不一致的影响,消除了上述算法中的颜色失真问题。
[Abstract]:The real-time scene and the range of luminance recognizable by the human eye far exceed the luminance range of the ordinary display device. In photography, if the brightness range in the real scene is large, the common shooting equipment can not capture all the details of the scene directly. The high Dynamic Range Imaging) method of high dynamic range imaging consists of hue mapping method and image fusion method. Compared with the method based on hue mapping, the method based on image fusion is more efficient and does not require additional exposure parameters. Since the image fusion method has been used to solve the problem of high dynamic range imaging and achieved satisfactory results, the problem of multi-exposure image fusion has been paid more and more attention by researchers. The multi-exposure image fusion method uses different exposure parameters to shoot the HDR scene many times. By fusing images with different dynamic range details, the details of different dynamic ranges in the scene can be preserved as much as possible. High dynamic range imaging is realized. In recent years, new image fusion methods have been used in high dynamic range imaging field. Empirical Wavelet transform (EWT) is a signal analysis and processing method proposed in recent years. It is an adaptive wavelet transform with empirical information. The empirical wavelet transform first analyzes the Fourier spectrum of the signal, then generates a set of empirical filters by a specific method. Finally, the generated empirical filter banks are used to filter the signal. Experimental results show that the empirical wavelet transform can extract the internal mode of the signal. Two-dimensional empirical wavelet transform (2D EWT) is a two-dimensional version of empirical wavelet transform, which can be used to process image data. There are many different kinds of two-dimensional empirical wavelet transform. In this paper, the simplest two-dimensional Zhang Liang empirical wavelet transform is used. This paper mainly discusses how to use empirical wavelet transform to deal with the problem of multi-exposure image fusion, and implements a multi-exposure image fusion method based on empirical wavelet transform. The luminance information of RGB image is related to three color channels of RGG and B. Brightness is an important reference in multi-exposure image fusion. If the three channels of RGB image are fused separately, color distortion will occur in some cases. To solve this problem, this paper proposes an improved multi-exposure image fusion method based on empirical wavelet transform in L*a*b color space. Because only the L* color channel contains luminance information in L*a*b color space, the algorithm avoids the inconsistent influence of luminance information on different color channels, and eliminates the color distortion problem in the above algorithms.
【学位授予单位】：吉林大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.41

【相似文献】