Right ! But, this does not mean the same result is obtained for all the images, as the setting is only a percentage applied to the map computed for each image.
Hmm, what i read is ... from within autopano, neutralhazer analyses each picture individualy, and decides to make different settings for each picture, and only a coarse adjustment from the setting the user choose from the one picture he himself decided to be representative?
There must be
... settings that say what is foreground, what is background, and how much they have to be cleared from haze. These settings should be the same for all pictures in one batch.
... a unique haze map for each picture, that (in combination with the settings) tells your algorythm how much each pixel must be cleared.
Hmm, the content of each pic is different, one contains more foreground, the other more background, so the haze map / foreground-background map will differ (in this point i agree). But all pictures should be filtered with the same "clearing settings". Otherwise the result will get too different.
Or is the way your dehazer works too different from other filters?
greetings from germany