Abstract
Existing multi-scale exposure fusion (MEF) algorithms which focus on defining weight maps to extract reliable information. However, they are difficult to preserve brightness order among over-exposed regions of a bright image and under-exposed regions of a dark image in a fused image when input images are two large-exposure-ratio images. In this paper, a new concept of exposure interpolation, which generates an appropriate bracketed sequence of virtual images from two input images, is introduced to address the problem. The possibly minimal number of virtual images is determined by using an optimization method. Then, an effective method by employing the intensity mapping function (IMF) is proposed to generate virtual images with proportional exposure times. The final image is produced by fusing both the two input images and the virtual images using multi-scale fusion. Experimental results show that the brightness order is preserved much better and the MEF-SSIM is significantly improved by the exposure interpolation.
Highlights
Due to limitations of current digital image sensors, it is not generally possible to capture the full high-dynamic-range (HDR) of a scene in a single exposure [1]
Methods based on deep learning have been proposed [5]–[9] which aim at restoring the lost information in over-exposed or underexposed images and generating an HDR image by using a deep convolutional neural network (CNN)
It is found that images with a large exposure ratio may yield brightness order reversal in the fused image if the input images are directly fused together
Summary
Due to limitations of current digital image sensors, it is not generally possible to capture the full high-dynamic-range (HDR) of a scene in a single exposure [1]. Their computational cost can be an issue for mobile devices Another common method is to fuse multi-exposed images into one image, which can well represent the shadow and high-light areas of the real scene. There do exist scenarios which need to reduce the number of exposures to simplify the hardware of sensors This leads to an interesting but challenging issue, i.e., only two images are captured, and the exposure ratio could be very large in order to cover wide dynamic range of the real scenes. Different from the idea of [33], the proposed method generates more than one virtual image to control the brightness order reversal, and theoretical justifications are provided to guide the exposure interpolation.
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
