[DONE] ML-based ND filter idea

Started by wuxiekeji, August 16, 2013, 06:06:18 PM

Previous topic - Next topic

0 Members and 1 Guest are viewing this topic.

wuxiekeji

Seeing that we can do dual ISO, can we implement an graduated ND filter by scanning different lines of an image at different ISOs (i.e. ISO 100 at the top, ISO 800 at the bottom) and then interpolating over the hard edges? This would be better than using a glass ND filter because we cut a layer of glass. No more colour casts, and hard/soft graduations could be implemented in software, saving a lot of money on expensive filters.

Also, this gets us graduated ND's for ultra-wide landscapes which isn't easy optically.

Dunc101

This would be incredibly useful. Not sure if it would look the same as am ND

3pointedit

Not sure that there would be enough ISO values to allow a smooth graduation? Just end up with banding?
550D on ML-roids

dmilligan

Quote from: 3pointedit on August 19, 2013, 12:26:31 AM
Not sure that there would be enough ISO values to allow a smooth graduation? Just end up with banding?

there are only two possible values you can use at one time since there only 2 ISO amplifiers, so making a smooth gradation would not be possible at all, I think.

I don't see why this would even be necessary, the dual ISO allows you to capture extra stops of dynamic range (which is really all that a GND does by simply assuming something about the lighting of your scene). You can simply implement your GND in post and you could make it perfectly smooth, and you can also put the transition wherever you want.