[C117] FlexMEF: Source-Aware Multi-Exposure Fusion with Controllable Brightness

Abstract

Multi-exposure image fusion commonly produces a single rendering, leaving subsequent brightness adjustment to post-processing of the fused image. Such adjustment no longer directly accesses the complementary input exposures. We present FlexMEF, which retains the short/long exposure pair for sourceaware reconstruction under continuous target-mean luminance control within a Base-dependent operating range. A monotonic tone reference establishes brightness, while source-conditioned tone allocation and detail refinement construct the proposed Fusion output. A Global rendering with shared detail corrections serves as an internal reference. On 50 SICE scenes, we evaluate six nonzero brightness targets—three darker and three brighter than the Base fusion image. At these targets, Fusion improves mean QAB/F and MEF-SSIM by 0.00293 and 0.00189 over brightness-matched guided low-frequency adjustment followed by a monotonic tone curve. Replacing downstream exposure inputs with duplicated Base fusion images under the same checkpoint reduces these metrics by 0.01354 and 0.00435 on average. A 101-target audit verifies numerical control fidelity, and Base quality is additionally evaluated on 100 official MEFB pairs. These results support retaining source exposures when controlling the brightness of a fused image.

Publication
IEEE International Conference on Electronics, Information, and Communication (ICEIC)
Jong Hwan Ko (고종환)
Jong Hwan Ko (고종환)
Associate Professor