An optimization-based approach to fusion of multi-exposure, low dynamic range images

The problem of compositing a high dynamic range (HDR) image for display on a standard low dynamic range device involves matte-based fusion of multiple images captured with different camera exposures, followed by a suitable tone mapping of the fused HDR image. The fused image should represent the entire scene in a clear, well-exposed manner by bringing the under- and over-exposed regions from the input images into the display range of the device while preserving the local contrast. We define matting as a multi-objective optimization problem based on these desired characteristics of the output, and provide the solution using an Euler-Lagrange technique. The proposed technique yields visually appealing fused images with a high value of contrast. Our technique produces the fused image of a low dynamic range, and thus it eliminates the need for generation of an intermediate HDR image and associated tone mapping. Additionally, our technique does not require any knowledge of the camera response functions or exposure settings.