- digital cameras have a limited dynamic range, which is lower than one encounters in the world
- in high dynamic range scenes, a picture will often turn out to be under- or overexposed
- Exposure Fusion has been proposed to skip the steps of tone mapping by computing a perceptual quality measure for each pixel which encodes desirable qualities
- assume images are perfectly aligned, possibly using registration algorithm
- for each pixel, combine the information from different measures (contrast, saturation, well-exposedness) into a scalar weight map using multiplication.
- weighting exponents being ω, i,j,k refers to pixel (i,j) in the k-th image
- normalize the values of the N weight maps such that they sum to one at each pixel (i,j)
- resulting image R can be obtained by a weighted blending of the input images I
# pseudo-code of Merten's Exposure Fusion Algorithm
C = scale(contrast(images))
S = scale(saturation(images))
W = scale(well-exposedness(images))
Quality_measure = computeWeightMap(C,S,W)
for each image I in the images do
pyr = Laplacian_pyr(I)
pyrG = Gaussian_pyr(Quality_measure)
for level L in the number of levels do
R[L] = pyr[L] * pyrG[L]
end for
end for