GH4 : Convert 4K 8bit to 1080p 10bit?

LPowell
In general, if you were to use the gh4 for a 1080 finish, would you shoot 1080 native, or 4K and "down-convert" to 1080? With quality of image being the main factor - as opposed to budget ….
 
Stuart U said:
if you were to use the gh4 for a 1080 finish, would you shoot 1080 native, or 4K and "down-convert" to 1080? With quality of image being the main factor - as opposed to budget ….
Click to expand...
With video, "quality of image" depends on what you're shooting. If I was shooting a static scene with only subtle camera or subject movements, I'd use 4K video mode to capture the finest moire-free detail, and downscale it to 1080p in post. But with panning or fast action, I'd be more concerned with motion picture quality than with static image quality. In that case I'd shoot directly in a 1080p mode with the maximum available bitrate.
 
Of course, the GH4 does up to 200 Mbps for 1080 - and 100 for 4K. I forgot that bit of info ( :
Thanks for the reply - and your wonderful posts regarding this camera. I look forward to using it.
 
For maximum image quality, it makes the most sense to always shoot in 4K 100Mbps unless there are visible macroblock/compression artifacts and/or rolling shutter issues. If disk/card space is an issue, shooting 1080p at 50Mbps has the 2x the bit-per-pixel rate as 4K at 100Mbps (almost- the higher res version could use optimizations for motion vectors applied to larger macroblocks. Shoot some tests to know for sure). If the artifacts can only be seen paused in an NLE and not while playing, 4K makes sense especially if the shots are a combination of motions (still, slow, fast etc.). I have been shooting exclusively in 4K with the GH4 (and delivering 4K to YouTube).
 
jcs said:
For maximum image quality, it makes the most sense to always shoot in 4K 100Mbps unless there are visible macroblock/compression artifacts and/or rolling shutter issues.
Click to expand...
Yes, that's exactly what I'm talking about. Macroblock artifacts are endemic to H.264 compression, but to what degree they're noticeable depends on the level of detail and motion blur you're dealing with, and also on how much grading latitude you need in post.
 
JanusCC said:
Did you follow my example about tone value "1019"? There is no way to represent that tone using 8-bit quantization.
Click to expand...

Sorry to bring this back onto the heated topic guys, but I'd been following along and this statement by Janus really confuses me because it feels like i missed something. I'm relatively laymen on DSP processes. But why did Janus say this? Isn't the point to represent 1019 in 10-bits? Of course we don't want to represent 1019 in 8-bits.

My understanding is that no one doubts that adding 4 8-bit numbers will make 1 10-bit number.
But people disagree on whether the resulting 10-bit number is "as good" as an "original"(?) 10-bit pixel.
I'm inclined to believe by virtue of it being able to hold a value that an 8-bit pixel cannot, that it is beneficial?
 
You must log in or register to reply here.
Back
Top