Messages

Hey there Steve - sure, thanks for asking.

You need to read up on 'stacking'. Its a common technique in astrophotography, where you place several frames (layers in potoshop), of the same scene on top of each other, register them exactly, and then average the pixel content through the layer stack in various ways. For instance Median or Mean values of each pixel stack. The result is that you get rid of random image noise and atmospheric haze, because these factors are random from image to image. I believe you need Photoshop CC to have access to Photoshops stacking functions, though it is possible to stack manually.

The more photos you stack, the cleaner the result generally. In this case, I used something like 240 frames off Magic Lantern Raw video. I could have shot 1000+ raw frames and chosen the best frames from that, but shooting video is way more practical in this scenario.

You can record in crop mode. You'll get 100% pixel readout.

Thanks guys! You should try it out :-) Its another of those computational photography tricks we love. In this case, raw video was just as good as using full res raw photos, since the MLV crop mode video corresponds to 1:1 crop on the sensor anyway. And the workflow was arguably easier using MLV (lower: data load, superflous data, shutter clicks).

Does your DNGs contain EXIF data? If so, try the lightroom or Camera Raw approach again, but use default noise reduction (or L+25 C=25).

Hi guys,

I thought you might enjoy this photo of the full moon. Its a 100% crop from a 400mm L lens on a canon 5DII.



I used MLV video to record it (crop mode on tripod), and then used about 240 aligned and averaged frames to achieve 'superresolution' - to eradicate random image noise, and obtain a much sharper image free of atmospherical disturbances. I did this in floating point mode to also achieve a higher color bit depth in PS CC (final image is 32bit, but also - no bigger than what you see here).

Best,

Mathias

>ACR will automatically remove the dead pixels.
>Doesn't seem to remove jack or shit... it makes them affect other pixels unless you turn the NR way up. What started as 1 turns into 4 at least.

Another case of missing EXIF info in the RAW files, I believe...

Why are you all benchmarking in photo mode?

It would be so helpful if a dev could set an override on all the small hacks when RAW video is enabled and youre in live view, then revert back to whatever was set before when you switch out of live view. Reason is, I'm a still photographer, and there's so many hacks and settings to set before RAW video (to enable large enough buffer and consistent fps), and these are easy to forget and easy to forget to revert when shooting stills. Its bothersome and dangerous. Something like a Custom Setting would work.

Just hot pixels, which is quite normal. Camera Raw has the ability to remove hot pixels automatically, and also do quality noise reduction, but for this to work, your .DNG files would need proper EXIF information, which they dont have for now. You can copy corresponding EXIF data (from a photo, for instance) into your DNG files via exiftool (search forum).

>when it can correct my lens profiles for vignetting

Right. Once you have copied the correct EXIF data into your DNGs.

I'd go with the 35 1.4 myself. Its not that wide, but the look of that lens is phenomenal. Do also consider Canons latest IS primes. IS can help quite a lot on handheld shots.

easy work arounds  that work just fine

But its not really an easy work around for most people, not least for people not aware, and having RAW footage with lots of noise to battle with in post actually negates having the data available for a real RAW developer to iron it out (with proper exif). Raw is just that. Raw. It really needs the exif container.

Eyeland, use high ISO and 3x crop mode. You should see a quite drastic difference from setting correct ISO exif value vs none. Also, make sure you are actually using the denoise sliders to put CR to work. And the 2012 version has much better denoising than 2010.

Is it just me, or does anyone else see a certain choppyness to the motion in the timelapses from ML ETTR timelapse? I see it in my own tests and in the other videos here. A certain constant rythm to the motion. Perhaps from varying exposure time?

Which values did you exactly change?

I'm not getting into exact tags. I'd go with what Muf and bnvm said.

Ah, the Fourier domain.

Full Exif would be wonderful, particularly I miss the ISO tag for Raw denoising.

I have this: https://plus.google.com/u/0/photos/102516342678541815125/albums/5898773988033658657/5898773988557612882?pid=5898773988557612882&oid=102516342678541815125&authkey=CIfd6p7T0KbiDQ

No exif is what we get right now. Exif is what you get when adding correct ISO information. Denoise settings are noted next to the image. Download and inspect color channels to really see the difference.

Best thing you could do to get less noise is lobby for exif information in ML Raw.

Camera Raw denoising is WAY superior to anything you can do later, since it draws on the full RAW data, and a database of noise profiles. Unfortunately, Camera Raw needs EXIF info to do its thing. Theres no Exif data in ML Raw video so far. I'm either using Neat Video, with much inferior results to what we would get with Exif in CR, or I use Exiftool to add the important ISO and camera make / model tags to the DNGs (so they denoise nicely in CR).

Probably (I'm guessing) Resolve uses some full-res proxy preview file embedded in the larger files (hence the difference).

You only see less noise in AE because the footage is darker / has a stronger curve profile. Looks the same to me. Problem is, there is no exif in the DNGs, so camera raw has problems removing the noise.

If we can capture 2144x1076 25 times a sec, and we can dolly this around with digital dolly, then in theory it would be possible to capture the whole raw frame at speeds like 1fps, but with some very odd shutter effects. If you could minimise these and combine it with fine exposure control and ETTR, you'd have a unique tool for timelapse.

Yeah. Or you could record at full 21 MP resolution and dolly in the NLE. Probably easier :-)

Hm ok. The exiftool route is very cumbersome. The thing is, Camera Raw has a database of noise profiles for various cameras. It needs the camera make and model for color profiles, but also the ISO speed for noise profiling etc.

I hope this demonstration kicked it up a bit on the priority list, as the difference is very big.

Thank you for all the hard work so far :-)

No problem - how do I post DNG's do you have a preferred upload site, or is it possible to embed attachments?

I'm maybe willing to eat my words on the 'always had exif' thing, but here's a quick comparison where I entered the correct ISO into a DNG via Exiftool:

https://plus.google.com/u/0/photos/102516342678541815125/albums/5898773988033658657/5898773988557612882?pid=5898773988557612882&oid=102516342678541815125&authkey=CIfd6p7T0KbiDQ (download or go full screen to inspect).

Huh? I've had exif in most all of my raw footage so far, except one earlier build. Now its gone again. I cannot. I stress. I cannot develop the DNGs without the EXIF! The raw developer doesnt know which camera, ISO speed etc it is! It looks like crap...