Exposure Fusion is a computer vision and photography technique that blends a sequence of bracketed photos (the same scene shot at different exposures) into a single, high-quality image. While “openExposureFusion” usually points generally to the open-source community’s various public implementations of this algorithm, the concept is famously rooted in a breakthrough 2007 method developed by Tom Mertens, Jan Kautz, and Frank Van Reeth.
Unlike traditional High Dynamic Range (HDR) imaging, exposure fusion completely bypasses the creation of an intermediate 32-bit HDR image. Instead, it directly generates a standard Low Dynamic Range (SDR/LDR) image by keeping only the “best” parts of each bracketed shot. How Exposure Fusion Works
The core algorithm evaluates each pixel in the image stack across three mathematical quality measures:
Contrast: Tracks high-frequency details (like sharp edges) using a Laplacian filter, prioritizing sharp areas over blurry or flat ones.
Saturation: Measures color vividness; pixels with rich, well-saturated colors get a higher rating than washed-out or heavily clipped pixels.
Well-Exposedness: Analyzes brightness using a Gaussian curve to favor mid-tone intensities. Pixels that are too close to pitch black (underexposed) or pure white (overexposed) are heavily penalized.
Leave a Reply