Multi-Exposure image fusion based on illumination map clustering

Limited dynamic range is a significant challenge in digital photography,as cameras often struggle to capture both bright and dark details simultaneously.This paper presents a novel multi-exposure image fusion method that addressesthis limitation by effectively combining multiple images captured at differentexposures. The proposed method uses the Retinex theory to decompose inputimages into reflection and illumination layers. A clustering algorithm is then appliedto the illumination layer to identify distinct illumination segments within the images.Contrast and well-exposedness metrics are employed to compute weight maps foreach segment and pixel, enabling a more accurate fusion process. The fusedimage is reconstructed by combining the weighted reflectance and illuminationlayers. Experimental results demonstrate the effectiveness of the proposed methodin enhancing the dynamic range and preserving image details.