Real-Time User-Guided Image Colorization with Learned Deep Priors (Aug 2017, SIGGRAPH)

Real-Time User-Guided Image Colorization with Learned Deep Priors
Richard Zhang*, Jun-Yan Zhu*, Phillip Isola, Xinyang Geng, Angela S. Lin, Tianhe Yu, Alexei A. Efros (*equal contribution)
In SIGGRAPH, 2017

[leonyp]

i want tutorial how to instal and use plz

[CBlasser]

I would pay you for a copy of this program as an executable, or even an install guide for windows.

[Frogmelon]

I've often wondered... For the original monochrome negatives - with appropriate scanning methods - could some kind of pattern (interference or otherwise) be ascertained at the quantum level, indicating the correct colour to use for a certain area. I'm just thinking that, even though the original does not preserve the colour information obvious to the human eye or to the original analogue techniques, perhaps the colour information is still available to be extracted, "hidden beneath the surface".

With old film, I've often wondered if colour information could somehow be retrievable from analysing the variations in luminance noise levels (for a particular pixel or group of pixels) that occur in a group of frames. At least to retrieve enough information to accurately tell us which colour to use for digital restorative techniques.

Edit: For instance: Perhaps for a given stock of film, "red" would be inherently more noisy than "blue", so that we can make an educated guess as to the correct colour.

Also, given that early cameras were hand-cranked so that exposure speed was inconsistent, perhaps the differences in exposure for each frame could introduce a colour noise signature which could be extrapolated once the precise exposure time for each frame was known. Just a thought.

[aloka]

Is there any AI colorization with a custom color scheme and interactive colorization,

because the automatic approach always have Color-Bleeding Effect