Greatings All,

I am curious about the resolutions > 1080p... I now have a gopro3 black that can record at 2.7K but what do I do with it??!! I can edit it in Kdenlive but, how can I render it in a format that will output at 2.7?

Is this a feature not yet available to video in general? Does anyone have experience with high resolution rendering?

Peace and love to all.

--HD

Higher resolution aquisition does not mean you have to output at that resolution.

It can be just as much about flexibility to recompose and zoom in a little, also benefit of scaling down to 1080P giving more detail than certain so called '1080P' shooting DSLR's that are actually far from it, for example.

So then it is no possible to render output to anything higher than 1080p?

Yes you can render greater thsn 1080P, but what is your target delivery and restrictions? Native resolution of playback device? Is that device going to scale it down anyway? With greater resolution comes more storage requirement, processing power and potential playback issues.

But you could creste a 2.7k project in kdenlive, edit and encode to h264 and play it on your PC, unless your screen resolution is greater than 2.7k it will be scaled. If that's the case maybe it's more beneficial to reframe, add some zooms, pans, scale and crop a more pristine 1080P video out the 2.7k?

Personally I think the oncoming push to greater than 1080P, in non cinema cameras for production, but then GoPros are used in production, is nonsense for anything but ability to reframe and crop unless I have a greater than 60" TV or projector in a home cinema set up other than going to cinema delivery. Its just marketing to get us to buy new TVs now 3D has failed.

Thanks for the reply Yellow. I just need some comfirmation of all this. I do see a difference in the 2.7K quality (or I think I do)...

So all this silly push by RED.com for the 'Dragon' and 'Epic' stuff pushing 6K/4K is just marketing until either new equipment that can display that rez comes out or more importantly, your eyes can see anything more than 2K... Anyway, I am still happy with my 1080p camera. Thanks again for the inside scoop.

It is also stupid how they changed the way of specifying format (resolution), just to make the numbers higher.

From the start of TV before WWII, it was always the number of scan lines (the vertical resolution), like the old GB BW system was 405, USA was 525, PAL was 625, old french BW was 819, etc. The HD specification still keeps the tradition with 480p, 720p, 1080i, etc.

BUT NOW - suddenly there are 2K, 4K etc formats!
So it would seem that 2K is twice the resolution of 1080? Noooo... they just started to use the horizontal resolution to name formats.
There is almost no difference between 1080 and 2k!
To make things worse, they can sell you all kinds of stuff under "2K", see

http://www.celco.com/formatresolutiontable4k.asp

@hounddogproductions, re RED, I wouldn't say it's silly and just marketing, our stills cameras shoot 4k raw, having the option to shoot video at that high resolution and raw or close to is great, for the reasons mentioned above and being able to extract high quality stills as if photos from the video stream has got to be good and it could be said that if it were'nt for RED other companies who now have to 'compete' with RED have had to up their game and deliver more for less.

All the 2k, 4k naming is because it's been aimed at Cinema, DCP an all that, different bodies creating the specifications. Different codecs and delivery. But the 'lines' are already becoming blurred (no pun intended) with the introduction of 4K for TV and home consumption.

A quick note on the 1520p and 1080p modes of the GoPro Hero 3 Black: as GoPro's Abe Kislevicz writes, the 1080p mode(s) are the 1520p mode with three fixed reframing and downscaling settings for wide, medium, and narrow. Instead of doing this in post production, the Hero 3 Black does it before writing the video data to the memory card.

Abe tested and found out that the results are virtually identical whether you manually post process or let the Hero 3 Black do the work. In consequence, you don't gain any additional benefits in terms of better sensor noise and light sensivity by switching to 1080p. It's always the same sensor, you are just scaling down. This also means that due to interpolation 1080p wide FOV will have the lowest noise levels (averaging is a low pass filter which typically reduces color noise), medium FOV comes next, and since 1080p narrow is almost 1:1 with sensor resolution, it will reduce noise the least.

That is also the reason why the Hero 3 Black needs the better sensor with reduced noise level, compared to the Hero 3 Silver and White. Otherwise, medium and narrow FOV would become unusable due to color noise. And Protune wouldn't work either, as noise reduction and sharpening is turned down when enabling Protune. You can see this with the Hero 2 where Protune basically becomes useless in certain lighting conditions and when you do your post processing in Kdenlive and avoid Cineform: the color noise is clearly visible and at least I never managed to reduce it using the Kdenlive filter without getting blurry results.

Lowpass reduces all kinds of noise, not just color, if done right.

Gopro's 2.7K is in fact 1440p (at 48fps! wow!) as specified here:

http://gopro.com/cameras/hd-hero3-black-edition

so it has cca double the number of pixels than 1080p. Thus, the spatial down sampling should give cca 3dB of noise reduction, which is just visible with the eye.

Of course, down sampling only reduces noise when done the right way, with a lowpass filter. Just using the "nearest neighbor" will not reduce noise, worse, it will add aliasing.

Whether it is better to downsample in cam or in post, depends on two things:

If the cam encodes 1440p with a higher bitrate than 1080p, then you get more info from the cam in 1440p, and it might make sense to downsample in post, especially if the cam uses "nearest neighbor" sampling.

On the other hand, if 1440 and 1080 are encoded at the same bitrate, it is better to leave downsampling to the cam, as the compression will be higher on 1440, giving no more info than 1080 - possibly just more artefacts.


@yellow >because it's been aimed at Cinema

yeah, I know. For the "cinema" guys, "video" is a dirty word, somehow automatically implying inferior quality, like "VHS is only good for porn..." - in fact for me, it's not even good enough for that :-)
So, they feel an inner urge to psychologically (and marketing-wise) distance them from anything "video".

But for an engineer like me, 2K (and 4K, 8k, 2^nK...) walks like a duck and quacks like a duck.... There is a 10X bigger difference between VHS and 1080p, which are both firmly in the video camp, than between 1080p and 2K.
Sure, XYZ is a slightly bigger (when working within an limited integer range) and better standardized color space, and higher bitrate coding means less artefacts, but it is all still just plain old video!

Denoising: I am still working on denoisers, sorry for the delay - you know that spare time is precious, and image denoising is a science in itself...
I've been experimenting with many different denoising algorithms, and it is not simple to decide which is the best. Probably no single one, as there are different types of noise, of base material, and different trade offs between speed and efficiency, etc.

Hope to have at least one ready before summer.

p.s. the existing denoiser is a port from Mplayer, where it was mainly intended for real time use during playback, so it's main advantage is speed.
Any advanced denoiser will be WAY SLOWER. (and yes, I am considering GPU acceleration if possible)

And of course, there is no denoiser that won't eat a bit of low contrast fine detail...

Small correction and nit picking: 1440p is actually termed 2.7k Cine in contrast to 2.7 which boils down to 1520p. I know for sure since I had to check a few things when I edited my 1520p footage taken during ice diving in Austria.

As for the Hero 3 Black: 1520p actually has higher bit rates compared to 1080p, you can find the max rates in GoPro's data sheet on their web site. So shooting in 1520p actually makes sense as there is room for post processing. Of course, 1520p doesn't always make sense.

And in my opinion, the integrated downscaler and denoiser is pretty damn good, at least for 1080p ... but unfortunately the white balancing is barely usable under water in many lighting conditions, except the best ones. So for us divers, we desperately need to shoot in Protune with wb basically switched off, which GoPro likes to call raw in a funny farm attempt.

The blog articles by Abe Kislevicz are really worth a read for GoPro Hero 3 Black users.

DiveO, I am curious as to what format you render to when shooting with the gopro3 black @ 2.7K (30fps). I have been using the Webm format but now I am starting to see a lot of degradation. Really I am just editing clips and putting it up on youtube etc.. but when I put stuff side by side (that is the original vs the webm render) it looks pretty watered down, e.g. mainly colors are off or washed out and such.

So I am wondering what format you render to.

Thanks in advance (all comments welcome) as I would like to know how people use kdenlive to render efficiently.


--HD

I have yet to work with the webm format. But independent of that one important factor is the maximum allowed bit rate. This is a tradeoff between size, detail, and transmission or storage capacity.


As a note, the 2.7k modes run at approx up to 45 MBit/s. That is 5.5 MBytes/s.

Final render output in my cases is good old 1080p25 H.264. This is mostly due to slow diving. For private use and archiving I currently render with 12 MByte/s and for YouTube I use 6 MByte/s. The rationale is that the infrastructure for high bitrates is still not there, at least where I live, despite all the computer illiterate politicians boasting otherwise. The 12 MByte/s are most probably overkill, but I got stick to it as I started with these para1in my early days.

But the other and more important reason is that under water in the waters I dive you have a lot of smallest debris floating which defeats highest resolution. So I use 2.7k to either scale down when I need the wide fov impression for a scene or scale down and reframe as necessary when I need the details to some degree. For the scenes I typically film 12 MByte/s already still has quite some reserve so that's why I'm sticking to it for the moment. But you need to check this.

This dive video is an example of shooting in 2.7k and in some clips I reframe and zoom in depending on my needs. The resulting video is mp4 container with H.264 codec and max bit rate of 8kbs.

http://youtu.be/SbQrYjc-yp8