Leaderboard

[media]http://vimeo.com/65260452[/media] These past days I have been trying to make myself useful to Alex at Magic Lantern and testing the DNG raw video recording on the 5D Mark III with a nightly build of Magic Lantern. Here's what the video quality looks like in the DNG video modes at 30fps. Click the image below to expand... [url=http://www.eoshd.com/content/10294/3-5k-canon-5d-mark-iii-raw-video-with-magic-lantern-and-latest-updates]Read the full article here[/url]

Poor Andrew, he gets so much shite for being a Canon hater, when deep down he really wanted to love the 5D3 for video. He's going all out, trying to save the 5D line from the evil engineers at Canon.   It's like watching someone perform CPR even though they know the patient is already dead.

    Absolutely, what videographer shooting hours of footage a day will process all that non-standard aspect ratio raw footage just for a bit of sharpness the client won't care about anyway for content that's probably web-only? And this is assuming the hack actually works...   I don't think most casual hobbyists would go through this trouble, either, which is why I don't think Canon is crippling anything video-wise (intentionally). Their debayering algorithm for stills (and video) is softer at 100% than Adobe's or DXOmark's and the in-camera processing throws away a lot of highlight detail. That soft algorith, alone might account for some of the difference in image quality between these raw video grabs and what the camera offers up as h264. Likewise I think peaking is useful, but it's very slow as ML implements it... None of the hacks are really production/casual hobbyist ready 100%. The C100/C300/C500 line is more a matter of crippling in increments, but dSLRs are still cameras first and foremost and the video is good enough for most.   I'm unsurprised that the 550d has the highest resolution of the previous line of dSLRs. Loved that camera!   Lastly, nearest neighbor downsampling has little to do with pixel binning as regards image quality. Nearest neighbor downsampling is after debayering is done. So you're downsampling all the interpolated data for each given channel. With pixel binning, every third pixel of a sub-1080p stream is red or blue with to weak an anti-aliasing filter to do anything. So the frequencies that alias are much lower than those that would alias in a downsampled still.   Because we lack the tools necessary to make really beautiful images on the cheap consistently (art design, lights, lenses, a big crew, talent) it's easy to get focused on image quality on the cheap and it is a valid pursuit. But it's kind of its own pursuit. I mean you can always rent an Alexa package for a week for about the price of a dSLR package.

A XF305 is totally different use to a DSLR.   Great AF, great zoom.   Also no pro with a C300 will dump that and go for a hack running on a DSLR just for raw or a better image.

I wonder which part of the "plan" was to lose both their pro/ama video film market to Black Magic?

Shot on the C100 with a Ninja 2. Prores HQ with a Canon 100mm f2.8 and a Rokinon 35mm T1.5 on C-Log. Graded with Color Finesse and Filmconvert. RGrain ultrafine:) Just trying to see how filmic i can get this bastard: https://vimeo.com/65086996

  Green color shift? Pinks to Orange as well? Nothing to do with WB. Use the right color matrix luma coefficients for BT601 instead of BT709 which will the assumed for HD source on pixel count if BT601 is not declared in the stream. Or other way round if Prores is flagging BT601 and the source is BT709.   Also is the source full range luma or not? Not sure if Prores even does full luma, but that's what the Canons output.

Well, I've been buying and selling on Ebay for around 10 or so years, but mostly vintage dinnerware, with the occasional electronics or camera sale when it was time to sell my old items.  The thing is, I rarely buy any dinnerware anymore there, since at least 50% of the time now it has undisclosed damage (chips, cracks, scuffs) that completely devalues the piece.  Unfortunately, all of the other collectors out there have figured this out too, so now if you are an actual GOOD seller, you can't get the correct value for your items since buyers won't pay premium prices for anything on Ebay (since they assume it will have hidden damage)....I've gone to selling my dinnerware to a private group of collectors through a message board and a couple of Facebook groups instead.  It's much nicer and easier all around, although it is a smaller audience than would be reached by Ebay (but oh so much less hassle...).   I've mostly had pretty good experiences selling electronics and camera stuff with a couple of notable exceptions, but the level of fraud that is going on now has gotten pretty high, and I can't really afford to risk that.  Ebay does NOT give even good sellers the benefit of the doubt, and that is a real problem. Buyers have learned they can beat the system and scam their way out of paying for their goods by using the "buyer protection" feature, and it's just gotten too iffy to sell there anymore (plus the cost is just insane now).   I have not sold the GH3 yet, though, so I may have to list it there at some point...I'll give it more time though... Thanks all for your comments!   -J

  I thought the buffer has a write speed limitation though, independent of the card?  I guess you'll find out if you keep sticking faster cards in there and performance plateaus.  

Thanks for the update Richard, i'll add the info to the topic start! @Pulp Writer: good stuff to know. So the lens is fixed focus and focussed at a pretty far distance I suppose? Personally I wouldn't spend $200 for this kind of results.

  I think the problem is doing a high quality debayer on the raw. The compression to JPEG isn't the processor intensive bit. It is the debayering.   All the camera has to do for DNG is dump the 1s and 0s to a very fast 512MB buffer memory chip. Then write 512MB to the card (that takes longer).   Imagine with every DNG file having to process all that data - 5MB of it - like opening a raw file in Photoshop - but in-camera.   Clearly the camera is doing a very simplistic debayer of the 2K raw to provide an image for live view and the H.264 encoder.   But we need a much higher quality one for decent 1080p MJPEG, otherwise you may as well stick with the existing H.264.   That's my understanding of it.   DNG is the option with the least processing, least CPU intensive, but sadly the most data to shuttle off to the card.   Of course Blackmagic cameras debayer their 2K raw just fine to ProRes in-camera, even the Pocket Cinema Camera. So it is doable with today's technology. What Blackmagic are doing could be very specialised though. Whether Canon have that technology, I don't know.

I am currently working on some concert footage I took in London last year. It was on my unhacked GH2 HBR 25p. Using my Canon 50mm lens set to F3.5/4. The ISO was 400 and I was using the Ex Tele mode. During the darker moments of the concert and in the shadows there is quite a bit of noise as expected. For my first attempt I used brightness/contrast & Shadow/highlight settings in PE 10 NLE. I them applied Neat Video. As I had a fixed camera position there is only one area that appears suitable for the software to use and after applying it I also added a second adjustment to the same area when the lighting increased. This added a bit to the profile but did not allow to fill in all the boxes on the Neat Video profile. I used the default settings. Overall the result is quite good but there is still a fair bit of noise movement in the darker areas. On a second attempt I also added the default Gamma correction to the mix which lightened the image quite a bit and also changed the Temporal filter radius to 2. However overall this increased the noisy area movement still further. I will be uploading my first attempt shortly. I am still using the demo version of Neat Video but as luck would have it the area it covers shows the most important part of the video for assessment. Are the adjustments I am using in PE brightness/contrast & Shadow/highlight and perhaps Gamma Correction the right ones to use? Any tips as to give the best base for Neat Video to work with?   http://www.eoshd.com/comments/topic/2195-laurie-anderson-dirt-day-concert-clip-testing-neat-video/ cheers!

  Yes it is single coated.