why 5k? go after uncompressed instead please

[Charles Angus]

Just thought I'd summarize what has been going on here. Seems the upshot is this (and I await correction).

1. More resolution at the same datarate gives better results.
2. The higher the datarate, the better the results - at the expense of filesize and ease of management. The obvious acme of this is uncompressed.
3. It seems to be a game of diminishing returns - that is, doubling a low datarate will provide a substantial increase in quality, but doubling a very high datarate would not yield a commensurate increase in quality. (A good example being SR - I defy anyone to tell the difference between SR and UC, even after a dub and cc).
4. There has simply been insufficient interest in an uncompressed option for Red to bother, regardless of the benefits.

 

[Charles Angus]

Speaking of SR, though, it would be nice to have a lightly compressed option. Something on the order of 2-7 to 1. Seems anyone satisfied with SR (and most everyone is) would be satisfied with this. Especially with the high quality Red ascribes to their compression.

 

[Dan Hudgins]

Yes, sort of...

Yes, sort of...

Just thought I'd summarize what has been going on here. Seems the upshot is this (and I await correction).

1. More resolution at the same datarate gives better results.

Well, going from 4K to 5K should not make the wavelet related artifacts at the same bit rate that much more noticeable, but if you went from 4K to 16K you would probably see more wavelet related artifacts (if you look at the images on a screen four times as big, if you use the same size screen your eyes may not be able to see the pixels anyway), tone banding and such. It depends on how much you low pass the image in the de-Bayer, if you go from 4K to 5K then low pass more to remove extra artifacts, then you might get some more shaprness in high contrast detail, but you lose contrast in low contrast detail and pick up chroma moire from the wavelets. The net gain or loss would depend on how the compression and de-Bayer are tuned to work together at that bit rate.

You do get an advantage going from 4K to 5K in that you can use a "lighter" OLPF and so when you resize to 4K or 2K you get closer to true resolution of RGB for each pixel. But blur from the OLPF and compressor is needed to control the bandwidth for a given bitrate, so if you have more pixels for the same bit rate you may need more OLPF blur to not overload the compressor with high contrast details.

2. The higher the datarate, the better the results - at the expense of filesize and ease of management. The obvious acme of this is uncompressed.

Uncompressed can be simple RAW or lossless "compressed" RAW. It would not make sence to use 3:1 compression with REDCODE if you could get 2.5:1 with lossless compression that give you 100% sensor data and less heat and power use in the camera. So REDCODE will probably never be used with ratios below 5:1.

The standard notion is that you cannot see the compression artifacts as much with high resolutions as low resolutions. That comes from viewing the same size image at different resolutions and compressions. The viewers eyes limit what you can see on the screen at a given distance, but what aspects of the image decay as the REDCODE compression goes higher may not relate to using JPG2000 ratios on graded images. I would assume that REDCODE would be tuned with the mating de-Bayer to put the losses in those parts of the image that would hide them best, that is the bit rate requires something to be lost, and how well you hide that loss is not a simple formula since what you see missing depends on the lighting of the subject and how close you sit from the screen (and if there is any subject or camera motion since some artifacts even if compressed just in the images can show up more when things move or the lighting shifts or changes).

3. It seems to be a game of diminishing returns - that is, doubling a low datarate will provide a substantial increase in quality, but doubling a very high datarate would not yield a commensurate increase in quality. (A good example being SR - I defy anyone to tell the difference between SR and UC, even after a dub and cc).

It depends on if you get closer to the screen or use a larger screen for the higher resolution. If 4K is far enough away not to see the pixels, then 5K with a higher compression ratio may not look much different except for maybe tone banding and "plastic skin" look.

If you are going to project IMAX, then you might see more advantage to a higher bit rate.

It would seem that going from 4K to 5K and reducing the compression from 10:1 down to maybe 6:1 would just about double the size of the R3D files, would the workflow at twice the file size be a true issue?

If you keep the R3D files the same size and go from 4K to 5K will anyone notice more artifacts on the screen projected the same width at a normal viewing distance?

4. There has simply been insufficient interest in an uncompressed option for Red to bother, regardless of the benefits.

Maybe for a 4K camera the filesizes are more than most would want to deal with. At 2K or 1920 with lossless compression the filesizes might be less of an issue, some of the fuzz people get using the camera at 2K comes from the compression, so having non-REDCODE for 2K or HD might make the images better, but it seems people would rather shoot 4K compressed then downrez to 2K or HD, so we will not get to make side by side images shot with the sensor true RAW vs. REDCODE to see exactly what losses REDCODE is making in the images... since there will not be a true RAW port (ever?).

I would like to see what could be done with the true RAW sensor data, and using true RAW sensor data for archive would let you pull out more detail and tone values in the future, REDCODE makes many filters on the sensor data that lose fine data that you can never get back.

For people who you work with compressed video, REDCODE is another codec on their list. Would making the images "better" make the "product" earn more, probably not. In the near future the RED cameras will not be compared to uncompressed FILM scans since they are on the way out, but to other compressed HD and Ultra-HD cameras. Will any other camera company try to use lossy compression BEFORE de-Bayer like RED has, or will they continue to de-Bayer in the camera before compression?

In otherwords does compression before de-Bayer offer a true advantage to the final result in the long run. It seemed to make sence when the camera was going to have a true RAW port to output both true RAW and compressed RAW so that the workflow would be the same for grading. But now that there will not be a true RAW output (?) why not de-Bayer in the camera like all the other HD cameras, you do not need to clip the files or do white balance you can make long scale in camera de-Bayer images. If you de-Bayer in the camera you can shift some of the chroma moire artifacts that the wavelets introduce out of the images.

 

[Dj Joofa]

In otherwords does compression before de-Bayer offer a true advantage to the final result in the long run.

I think you do have a point as the compression done before deBayer would result in compression noise smeared into neighboring pixels by the deBayer process.

 

[Joseph Ward]

What would uncompressed RAW be for 617? lol!

 

[Deanan]

What would uncompressed RAW be for 617? lol!

Compression efficiency goes up and up as you increase resolution.
Resolution also has very positive effects on usable latitude as it goes up.

 

[Gavin Greenwalt]

it wasn't so long ago everyone was happy with hvx 200 and that footage is compressed like crazy..

I don't remember agreeing to this.

 

[Gavin Greenwalt]

It would be a much better archive format if Red would release tools to allow for trimmed R3d files.

Well there was this one too~./?+^###S###...

*Connection Lost*