Tests at different filter positions show strong assymmetrical vignetting, and blue colour cast in the darkened areas for the highest attenuation positions:
The vignetting pattern changes quickly from one position of the filter to other, which makes imposible to obtain a standard correction map. I have done a numerical correction for the 13.5 position, and worked quite OK eliminating both the darkening so as most of the blue cast:
Vignetting and colour cast (obtained taking a picture of a flat surface), non corrected image and corrected image (colour banding was caused by rounding errors since the calculations were applied in 8-bit):
The R correction code, just four lines:
# Leemos imagen
vignet=readPNG("vignet.png")
# Normalizamos
for (i in 1:3) vignet[,,i]=vignet[,,i]/max(vignet[,,i])
# Aplicamos corrección a la escena y guardamos normalizando
# OJO: ningun píxel de vignet en ningún canal puede valer 0
scene.out <- readPNG("scene.png")/vignet
writePNG(scene.out/max(scene.out), 'scene.out.png')
Regards