HDR Still Editing

[sanlyn]

Right, poisondeathray, I get the general idea. For some reason this masking/blending is a slow study for me. As for the VHS project, it's a matter of compressing the range, not extending it; you know how some VHS plays with outta sight levels and whatnot. Denoising isn't a problem, it's the screwed up histograms I'm looking at. The usual color controls get me only so far. Maybe it's now the best it'll ever be, but it keeps nagging me. I spent all that $$$ on Adobe and just built a new PC for it, I might as well dive in and explore different avenues.

[poisondeathray]

Right, poisondeathray, I get the general idea. For some reason this masking/blending is a slow study for me. As for the VHS project, it's a matter of compressing the range, not extending it; you know how some VHS plays with outta sight levels and whatnot. Denoising isn't a problem, it's the screwed up histograms I'm looking at. The usual color controls get me only so far. Maybe it's now the best it'll ever be, but it keeps nagging me. I spent all that $$$ on Adobe and just built a new PC for it, I might as well dive in and explore different avenues.

Well there is only so much you can do with VHS....

For this topic in general (not tone mapping with different exposures), it's the same way you would make "mask selections" in something like photoshop . You don't necessarily have to manually draw a mask or rotowork

You ask yourself: "how to I isolate this?" You can isolate by various techniques, say a range of luminance, a certain range of hues (really, this is the basis for secondary color corrrection), maybe chroma keying . You combine them by compositing layers and masks. Many of the same approaches have something similar in avisynth, but after effects is way more versatile for mask work. The last resort is manual masks

[sanlyn]

I understand. Yeah, I'm down to manual masking. But I do have 9 captures of this video, each would be something like the equivalent of different exposures (well, more or less). No one capture looks like the other. Some flaws are common to all 9, some are not. This is definitely going to take time and study. Likely you're familiar with the tape from another thread of mine. It's infamous as one of the worst film-to-VHS transfers ever. But I'm making headway. Some of the ideas posted here would be worth a look.

[poisondeathray]

I understand. Yeah, I'm down to manual masking. But I do have 9 captures of this video, each would be something like the equivalent of different exposures (well, more or less). No one capture looks like the other. Some flaws are common to all 9, some are not. This is definitely going to take time and study. Likely you're familiar with the tape from another thread of mine. It's infamous as one of the worst film-to-VHS transfers ever. But I'm making headway. Some of the ideas posted here would be worth a look.

Were they 9 captures from the same VHS source at different settings , or were they different broadcasts , period?

In the former, it will be the same MPEG2 8-bit broadcast source from the TV station at the same exposure, so it wont matter. Only if you did something during capture to clip or crush values in one of the instances would it make a difference. Changing the "exposure" during capture of the same source doesn't matter otherwise. You could do the same thing in software as long as no values were clipped or crushed

In the latter, it might be possible to make improvements if they were different versions

Was this the "Lili" thread ? Then you might be able to combine elements using masks and compositing since there were different sources like laserdisc .

[sanlyn]

The same VHS source, different settings, 4 VCR's. That's LILI, yep. Every camera shot is wildly divergent in every respect from surrounding shots. You always get that with VHS, but this movie is ridiculous. It's my 3rd copy of the issue. Out of all those copies, I couldn't get an intact 2 minutes of opening titles. so I borrowed them from a broadcast. The broadcast is otherwise horrible. The 9 capture settings were intended to get a collection of captures wherein each would yield some workable footage. It's odd that the "best" capture is #9, from the 1995, oldest and cheapest of the players. Who would have guessed?

[budwzr]

S.lyn@ hey, wtf, your not cohesive. Are you flashing back to nam?

[sanlyn]

What's up, budwzr? Sorry for the diversion, we went from applying HDR in different ways and somehow tripped into something that belongs in another thread. I think HDR theory might be useful in that video, but I'll save it for the other thread.

[budwzr]

Hey, look. Theres a bird.

[sanlyn]

And that's what my comments on the various shadow/highlight/midtones filters do - the base their corrections based on predetermined level ranges

You can even do this in avisynth with the masks pack. e.g. isolate between Y=0-40, 41-130, 131-255 (or whatever range, etc...) , then apply your filters through those masks .

My point was you can't extend the range and detail that is already there from the 1 exposure. e.g. when you brighten shadows, you often reveal crap and noise. By using real tone mapping techniques, and merging a more brightly exposed "shadow" version, you can effectively create an image that has nice shadow detail without the noise and need for heavy denoising

Yes, that's what I'm hoping to accomplish with the different captures (i.e, different "exposures", more or less). You explain it far better than I do.

[budwzr]

The Magic Lantern guys with the Canon T2i have this in feature programmed in the firmware where they take alternate exposures at 1/2 the framerate (but this way you don't need a special rig and different cameras to take the shots). The extra "inbetween" frames are synthesized in avisynth or similar techniques - it only works well on certain types of shots. You can see some great examples on vimeo and youtube

What do I search for? HDR Video?

[poisondeathray]

The Magic Lantern guys with the Canon T2i have this in feature programmed in the firmware where they take alternate exposures at 1/2 the framerate (but this way you don't need a special rig and different cameras to take the shots). The extra "inbetween" frames are synthesized in avisynth or similar techniques - it only works well on certain types of shots. You can see some great examples on vimeo and youtube

What do I search for? HDR Video?

LOL.

Grade: F- for Internet Search ability

Search for Magic Lantern Firmware, HDR . Software is used to interplate frames (optical flow - a free way is avisynth with mvtools2) , and software is used to for tone mapping video e.g. GingerHDR for after effects (but there are many other options)

[aedipuss]

here are the hdr video samples. i use m.l. 2.3 in my 60d but haven't tried the hdr video yet, the rest is nice.

http://www.magiclantern.fm/hdrvideos

[Cornucopia]

The problem I see with ML's HDR video is that it:

This feature allows you to shoot a high dynamic range video by alternating ISO every other frame.

Which means you are getting frame-blending effects, even if it's using motion-estimated processing. I wouldn't want that.

I've come up with a rig that works similarly to the way Stereo3D Beamsplitter rigs do, except the difference is not in placement, but in exposure. Yes, there is slight double-image and color-cast differences between the 2 perspectives (due to the beamsplitter) that need to be corrected/matched, but at least you get FULL, clean HDR frames at the full framerate.

Scott

psst:
(I also have another rig that works even better, doesn't have even those problems, but can't fully divulge right now as I'm hoping to apply for a patent on it)

[poisondeathray]

The problem I see with ML's HDR video is that it:

This feature allows you to shoot a high dynamic range video by alternating ISO every other frame.

Which means you are getting frame-blending effects, even if it's using motion-estimated processing. I wouldn't want that.

Yes, but is a "cheap" way to get it. This is why I said earlier only works well for certain type shots. Certain types of shots are more predisposed to the morphing artifacts you get from interpolation .

Even more expensive, full framerate, in camera solutions like Red Epic's HDRx have problems since the exposures are taken at different shutter speeds - they will have different amounts of motion blur

psst:
(I also have another rig that works even better, doesn't have even those problems, but can't fully divulge right now as I'm hoping to apply for a patent on it)

Nice! Maybe you can give videohelp members a discount when i comes out

[budwzr]

here are the hdr video samples. i use m.l. 2.3 in my 60d but haven't tried the hdr video yet, the rest is nice.

http://www.magiclantern.fm/hdrvideos

Whew! Nice!

I gotta get me one of them fluid heads one of these days.

P.S. ML is 2.3 now? Time to upgrade, but I'm scared to make any change, lest something go awry.

[aedipuss]

haven't heard of any upgrade problems. but i'd use the update firmware option to remove the old ml completely first then install a clean 2.3

[aedipuss]

i'd like a cam that shot hdr video at 72p high/normal/low and combined it into 24p - please? even 60p high/low might be acceptable, but missing the "correct" exposure has drawbacks.

[Versus]

For this price you can have a better option, or even find an HDR software with a bundle of special offers. Adobe is dubious, you'd better buy Aurora HDR https://aurorahdr.com/ for the same price.