I have done some side by side comparisons of clips shot at 4K 100 Mbps and 1080p 24 Mbps and I can't see any difference if any sort between the two.
Furthermore I have compared the down scaled footage with the LX100 and the avchd on the GX7 with kit lens and the latter looks actually better
What are the theoretical benefits of downscaled footage as I can't seem to see any???
The story goes that if you do the downscaling the right way, you can effectively have 4:4:4 colour sampling to work with instead of 4:2:0, and debatably you can increase bit depth from 8 to 10. A visual comparison on your monitors probably wouldn't reveal that extra quality, but you might be able to do more grading etc. with it in post if you do it the right way. Lots out there on this subject including this.
Also, you have scope in some methods to use different downscaling kernels such as bilnear, bicubic, Lanczos and spline. These affect sharpness and give you some element of control in the look when you downscale.
Interesting post. But the question is; wouldn't you get similar results by color grading first and downscale the master rather than downscale every 4K clip to 1080 an then do the color grading in a RGB 10bit project?
Am not able to use an external recorder the file is 4:2:0 8 bits and actually grading is not my concern. Am purely talking about resolution in scaling. Whatever algorithm is in the camera looks like it achieves the same results or better. So is the pixel binning in camera better than the scaling of the video editor? Am specifically interested in Mac products like final cut as I don't have a PC
From a GH4 perspective.
The Internal 1080p is much softer vs taking a 4k file and putting in on a 1080p project timeline, the same is also true for some reason when getting the GH4 to downsample the 4k to 1080p via the HDMI port.
I think its due to the internal 1080p mode taking the whole width of the sensor and squeezing it to 1080p, where as the 4k > 1080 its a straight 4 pixels into 1 pixel downscale.
From a GH4 perspective.
The Internal 1080p is much softer vs taking a 4k file and putting in on a 1080p project timeline, the same is also true for some reason when getting the GH4 to downsample the 4k to 1080p via the HDMI port.
I think its due to the internal 1080p mode taking the whole width of the sensor and squeezing it to 1080p, where as the 4k > 1080 its a straight 4 pixels into 1 pixel downscale.
I have only an LX100 what you are saying is that with a 1:1 pixel ratio you can then do a linear scaling (2:1) and it looks better than 4608 to 1920 that is 2.4:1
That seems to be a benefit in the reduction of artefacts however the LX100 doesn't take 1:1 pixels and operates from 4112 to 3840 on the horizontal axis. The resolution of the camera itself looks lower so when I compare it to an 1080p clip it doesn't look much different
Also the 1080p of the GX7 straight from the camera looks better
It appears that unless I produce a 4K clip with the LX100 downscaling seems not to be worth it
I wonder what's the status with the GX8 that also doesn't take pixels 1:1?
Aside from sharper or softer results downsampling, I wonder if the best workflow for color grading for a final output in Full HD is:
-Shoot in 4K, work de editing and grading in a 4k project and render and mastering downsampling to Full HD
or
-Shoot in 4K, down sampling all the clips to Full HD 422 10 bits, and work the editing and grading in a HD project with the new clips
My choice would be shoot and edit in 4K and deliver in 1080.
With FCPX I would also transcode to ProRes, so the system isn't having to transcode on the fly. Working with the higher resolution images will provide more scope for color correction, stabalisation etc, plus noise reduction applied to a 4k file and then down sample really does clean up the noise.
The only downside to all this is that we (at the moment) lack 4k 60fps and we have to live with sensor crops so taking away some of the wide angle benefits of our lenses.
... however the LX100 doesn't take 1:1 pixels and operates from 4112 to 3840 on the horizontal axis.
I think this is why you don't see any benefit from downscaling the 4K to 1080p - as the 4K mode from the LX100 is already performing some pixel squeezing.
I think its only the GH4 that is taking a pure 1:1 4k centre crop from the sensor.
The GX8 sensor has a greater crop (due to being 20MP) in 4K mode - and I think (from cameralabs review) its 4K mode is also not a pure 1:1 pixel reading
I wish they would give use the options
1. Use full width of sensor for 4k - but loose quality
2. Use 1:1 centre pixels - but lose some wide angle
it may
I think this is why you don't see any benefit from downscaling the 4K to 1080p - as the 4K mode from the LX100 is already performing some pixel squeezing.
I think its only the GH4 that is taking a pure 1:1 4k centre crop from the sensor.
The GX8 sensor has a greater crop (due to being 20MP) in 4K mode - and I think (from cameralabs review) its 4K mode is also not a pure 1:1 pixel reading
I wish they would give use the options
1. Use full width of sensor for 4k - but loose quality
2. Use 1:1 centre pixels - but lose some wide angle
i think it has also to do with the fact that the LX100 has limited resolution
The GH4 resolves 2400-2500 lpph the LX100 only 2150 so in effect it does not even resolve 4K at the start
The 1:1 pixel mode is very useful for macro on my GX7 it gives an immediate 2.4x zoom and it looks very clean
My choice would be shoot and edit in 4K and deliver in 1080.
With FCPX I would also transcode to ProRes, so the system isn't having to transcode on the fly. Working with the higher resolution images will provide more scope for color correction, stabalisation etc, plus noise reduction applied to a 4k file and then down sample really does clean up the noise.
The only downside to all this is that we (at the moment) lack 4k 60fps and we have to live with sensor crops so taking away some of the wide angle benefits of our lenses.
I looked in FCPX and it does not have options in the scale function to determine how it will be performed. I understand the Prores part to make changes however I still don't believe it affects scaling.
It would be different if the footage had been captured in prores to start with at higher bitrate and then certainly there would have been less artifacts and more detail however this only applies if you have an external recorder.
Other than the Canon 5D, Panasonic GH4 and Sony RX100 and A7 am not aware of other devices that can use an external recorder?
Other than the Canon 5D, Panasonic GH4 and Sony RX100 and A7 am not aware of other devices that can use an external recorder?