Research Of Multi-exposure Image Fusion Based On Empirical Wavelet Transform

The range of luminance of a real scene and that can be recognized by human beings is far larger than the common display device.If the range of luminance of the real scene is very large,common cameras can't capture all the details of the scene when photographing.High Dynamic Range Imaging(HDRI)methods can be categorized into two main groups: tone mapping based methods and image fusion based methods.Compared with tone mapping based methods,image fusion based methods need less computation,and don't require exposure settings.Since image fusion based methods were used in HDRI and performed satisfactory effects,multi-exposure image fusion has attracted more and more attention from researchers.Multi-exposure image fusion methods take multiple pictures of the HDR scene with different exposure settings,to preserve as much details as possible of the scene by fusing those images which contain details of different dynamic range,realizing the goal of HDRI.More and more image fusion based methods were used in HDRI recently.Empirical Wavelet Transform(EWT)is a newly proposed method of signal analysis and processing in recent years,it is an adaptive Wavelet Transform(WT)with empirical information.In EWT,the Fourier spectrum of the signal is analyzed,and a set of empirical filters are generated,then those empirical filters are used to filter the source signal.It has been proved that the inner modes of the signal can be extracted with EWT.2D EWT is the 2D version of EWT which can be used to process images.There are many different 2D EWT methods,and 2D Tensor EWT is used in our experiments.This paper mainly discusses the subject of how to apply EWT to solve the problem of multi-exposure image fusion,and a EWT based multi-exposure image fusion method is realized.The luminance of RGB images is related to all the three color channels of the images,and luminance is a very important reference point in multi-exposure image fusion.If three color channels of RGB images are processed with EWT separately,there will be color distortion issues in some fusion results.Another improved EWT based multi-exposure image fusion method which performs the transform on L*a*b color space is proposed to eliminate color distortion in fusion results of the previous method.Only L* channel contains luminance information in L*a*b color space,the new method avoids the different effects of fusion in different color channels,and eliminates color distortion.