Messages

This will be very helpful, big thanks.
So, this is on Half-res blending mode? Any chance to apply blending full-res? 0k, that requires Octave, but it could be done. Moreover, as mentioned above, perhaps the new features of 20bit version and "film-curve" could help.
I will test soon and bring you the feedback. Also, try to do the translation into Portuguese - BR.

+1
And the sRAW buffer method... would be very useful.

[@a1ex]

@a1ex

What edge-detection method cr2hdr use? EECI? I find one very interesting paper, of three koreans, in proposing the new method of edge detection method and the refinement scheme, called "DVBDM" or "Directional Variances Based Demosaicing Method". See here.
The algorithm described is 37,8% more faster than EECI (avarage), and have more 2.95 dB of result on Color PSNR measurement.
I was wondering if it would be possible to adapt something to use in cr2hdr... but unfortunately I am not mathematician to do it.

*only problem of that, which is used is 5x5 CS.

600D 

I have not used rawtherapee. What is it for?

It is a RAW engine, like a ACR, but open source and better: https://www.rawtherapee.com

[@ted ramasola]

@ted ramasola

You can do it?

Open this chat image on RawTherapee and post some samples with LMMSE and IGV algorithm?
I saw this same effect in the lower right corner on other images, and generally these two interpolation artifacts came. This would explain the adcional sharpen, because these two algorithms are optimized for high noise.

As far as I know, Adobe uses its own modification of AHD for demosaicing. I looked quickly on the BM algorithm and found nothing (anyone there?)... but it is likely that they use an algorithm less complex because of application in video, which would explain the pronounced aliasing.

Well this would be the most "correct" process, so to speak. However, you can work at ACR, only changing the color profile in which he is working. Try using ProPhoto RGB or any profile designed for film, so you get all the information.
The grain or texture, always apply as the last layer of all processes, after color grade, denoise/sharpen... after all done.

Hi guys,

[...] Use a grain image (like rGrain stocks) and apply to the composition  -  Export in uncompressed mode (is this correct?!)

3- Inport this videos to Davinci resolve and, because you are working in the BMDFilm Color space - apply the BMDFilm-Visioncolor LOG lut  - than do the colour work and transform things like you want to. [...]

So, I use this workflow too, but, I believe there are two points where you went wrong in your flow: application of grain and color space.
Every type of texture is applied after the color grade, otherwise it will have a negative effect on texture.
What about the color space, you must work with the same color space throughout the flow. Note that you will initially use sRGB (set by defaut on AE), which is a very small space, and is subsequently you convert this to a larger profile. This way you are not taking advantage of all the information, on the contrary, after the first render you ruled out all this information.
Try to work the entire process from the BM space or work in ProPhoto RGB since the beginning...

Level 5.2 is not supported? And, it's config. is ALL-I [GOP 1]? Maybe this instability is because you set very high maxbitrate... try to config. less, and analyse it with appropriate tools (push to the limit curves, for example, this will show how different macroblocks stay).

Ah, yes, if your card is not fast enough, you can create dng that's magenta contaminated... these frames are corrupted and become unusable. Besides the magenta frames, as aforesaid, there is the so-called "bad pixel" where there is no information and that end up generating a colorful points account debayer interpolating.
Escho created a compilation of raw2dng which prevents the latter from happening.

Greetings.

[...]jumps back to the photo screen instead of start to recording. What is the problem?

Are you sure you're in video mode? You install the official 2.3 version or Nightly Build?

But the article say "greatly improve dynamic range" 
Best quality low-pass filter can provide more DR?
My guess is there may be some physical change with respect to the input signal. The problem is the way they sell this product... that it is not in every city there is a Canon authorized store.

[this]

See this topic.

I use Cineform to transcode... its work smooth to me. Workflow: Escho raw2dng compile > LR 5 > TIFF sequence on AE CC > Cineform 444 > Video Enhancer (resize) > PP CC.

Exp. the Cineform or Lagarith on your software and use, for example, Video Enhancer

MeGUI replacing the x264 for the current, and replacing the qaac too.

Try changing the FPS of its sequence. Right click on your sequence of DNG button and go to "Interpret footage". Change from 30fps to 23,976 fps.
Moreover, it may be that your project is in another fps. In this case, do the same but with the icon of the project, change to 23,976 fps.
If the problem persists, it is likely that you do not have enough processing power to play the upscale real-time video. Render ("RAM preview" on AE or ENTER on PP), and then yes you can do it.

Report your bugs for dev. of RT project, please. Seen's normal to me here...

No problem here, too. Try update rawtherapee and install new builds of ML.

Can up one dng?

Can up. some sample?

I usually use MeGUI,  replacing the x264 10bit to it. I export using CRF 24, preset Film and others advanced option. But, anyway, youtube not like quality and will end up with your video the same way.

[ah, I use Cineform to raw export]

 

Thanks for the responses guys.

Continuing your comments... "And new studies are declared to be so promising that it is now possible to create images almost noiselessly, with peak SNR never seen before. Quantum physics applied to software, perhaps? Are only a1ex and Marsu42 know and say it." He continues: "It seems that no limits, 4K RAW at 60fps, non-cropped via HDMI out!".
So funny your comments, please continue, are really good; hahaha.

Don't hear that much complies about JPEG NR in stills.
Conclusion: You're thinking about RAW video capture converted to h.264 or else by ML?

On middle process, maybe a interference by the ML...