Patch-Based Approach for the Fusion of Low-Light Image Pairs

When taking pictures in a low-light scene with artificial lighting, we often encounter the following problem: we can use short exposure setting which yields a dim, noise image but with a sharp outline, or we can use a longer exposure setting which yields a bright, saturated image but with blurred areas. In many cases, none of those images is good enough. Good brightness and color information are retained in longer-exposure images, whereas sharp outlines are retained in shorter ones. In this paper, we present a patch-based method to combine such image pair into a better one. In our method, we firstly perform a coarse-to-fine strategy to detect inconsistent pixels caused by moving objects, then we draw information from the two exposures based on a novel patch-based technique. Experimental results show that the proposed method effectively preserves sharp edges of the short-exposure image, and maintains the color, brightness, and details of the long-exposure image.