Welcome to our community

Be a part of something great, join today!

Log in
  • Hey all, just changed over the backend after 15 years I figured time to give it a bit of an update, its probably gonna be a bit weird for most of you and i am sure there is a few bugs to work out but it should kinda work the same as before... hopefully :)

How to test the actual visible differences in compression ratios

TJ Hellmuth

TJ Hellmuth

Well-known member
Joined
Jan 29, 2009
Messages
110
Reaction score
0
Points
0
Age
47
Location
Santa Monica
Website
www.tjhellmuth.com
I'm working with a photographer on some television commercials, mostly for fashion and beauty brands. Coming from the photography world, he's not used to thinking of his raw files having compression, and to him, he'd theoretically like to shoot 1:1 ratio because he simply wants the best possible quality. So he'd like to do some tests to actually examine the differences between compression ratios. It would also be great to know exactly we are compromising when shooting at higher frame rates and HDR. i've read lots of rules of thumb, but also seen very conflicting practices in this regard. Some DP's will not budge from 5:1 whatsoever. other people on the forums say that 8:1 is a good sweet spot for non-effects work. so I'd love to see the difference for myself.

Is there an example online of anyone looking at just how far they can push these files and what exactly you loose visibly when you shoot at higher compression rates? I'd like to form a test for him, and I'd love to post the results here but I'm trying to figure out how to test. What will the weak points be that I should focus on? color depth? subtleties of skin tone? range? Sharpness? Any links or ideas for a good test would be appreciated.


Thanks!

TJ
 
Depending on how r3d compresses a frame, if you create a small 3" section of ultra fine detail, like a piece of a mft chart, then for the rest of the frame have a ton of fine detail moving, like filling the frame with moving water, and shoot the various compressions. You should see the compression in the chip chart.


Again this depends on if r3d compresses each frame independently, regardless of content. Also, a range of contrast would probably be good.

I see a lot of artifacting in 12:1, but almost none in 10:1, so it seems there is a sweet spot where the bandwidth can overcome the details.
 
Thanks for the reply. This sounds like a good plan. Perhaps a shiny colorful pinwheel might be a good alternate to the water in terms of rigging and range of color and latitude. This sounds like a good test. Thanks!
 
Timur Civan said:
I see a lot of artifacting in 12:1, but almost none in 10:1, so it seems there is a sweet spot where the bandwidth can overcome the details.
Click to expand...

When shooting for a 1080p post and a finish possibly smaller then that (720p/SD) are you content with shooting 10:1 to save space? I'm talking straight cut and output; no compositing, green screen, crazy color grading.
 
TJ Hellmuth said:
Thanks for the reply. This sounds like a good plan. Perhaps a shiny colorful pinwheel might be a good alternate to the water in terms of rigging and range of color and latitude. This sounds like a good test. Thanks!
Click to expand...

The secret to stressing the codec is detail. The pinwheel is likely to create motion blur, which actually makes it easier for the codec, because it has less detail (more blur). Deep focus shots of water, as Timur suggests, or gently moving leaves are ideal stressors. Lots of fine, sharp detail=significant challenge for a codec.
 
Also, try to pull an aggressive HSL qualified secondary grade - any compression noise will pop right to attention.
 
The compression Red is using in R3D's does not get effected by moving objects, it does not use motion vectors like mpeg or h264.
So a tree moving in the wind is no harder for it than a static tree.
Lots of random detail is stressful. Gamma up the image and look at things like tree bark.
 
William Robinette said:
When shooting for a 1080p post and a finish possibly smaller then that (720p/SD) are you content with shooting 10:1 to save space? I'm talking straight cut and output; no compositing, green screen, crazy color grading.
Click to expand...

I do 10:1 for doc that isnt going through a DI and is just being bumped down to Prores for a quick edit up to You tube.

anyhting that has a chance of being graded is shot 5:1, 3:1 for VFX shots.
 
5:1 or 6:1 is what I usually stick with unless it's shots with a lot of detail like wide, deep focus shots of trees- then I'll shoot 3:1. If storage space or offload times are a concern I'll shoot 8:1 or 7:1.
 
Thanks guys. This sounds like a good amount to bite into. Probably explains why I had some weird noise when I was shooting high speed of the wind swept ocean recently. it was all in focus and all moving slowly and tons of glistening detail. This is great info. Thanks!
 
Visible is so subjective. Shoot something. Look at it on screen. I don't see how doing a bunch of technical tests changes what your eyes see, or think they see.
 
Chris, I'm not sure I get what you are recommending. Shoot, but don't make it a technical test? The purpose of what I'm doing is to give tests to a director/DP so he can have more confidence in what ratios he shoots at and can know what he's willing to compromise in favor of frame rates or HDR. So I'm not wanting specifically to keep these tests technical but I want to perform one set of tests that gives as definitive an answer on this topic as possible in the least amount of time or error or theoretical guessing. So these test ideas like bark , water, trees and charts are great ideas.
 
TJ Hellmuth said:
Chris, I'm not sure I get what you are recommending. Shoot, but don't make it a technical test? The purpose of what I'm doing is to give tests to a director/DP so he can have more confidence in what ratios he shoots at and can know what he's willing to compromise in favor of frame rates or HDR. So I'm not wanting specifically to keep these tests technical but I want to perform one set of tests that gives as definitive an answer on this topic as possible in the least amount of time or error or theoretical guessing. So these test ideas like bark , water, trees and charts are great ideas.
Click to expand...
This might be a good rule of thumb:

Pixel peeper (like myself) or VFX = 3:1
High quality general footage = 5:1
Documentary = 8:1
Talking heads = 10:1

Thoughts?
 
A lot of talks have been going on for the Panasonic GH2 Hack (I know DSLR...)
The hack is about enhancing the bitrate and therefore gaining details.
What has been used is what they call a Death Chart (where camera codecs goes to die)
you can get it here: http://www.personal-view.com/talks/uploads/FileUpload/b2/79e08da4dae9d445734b214f4fa06a.jpg
Print it in large and film it handheld (for some extra jitter to stress the compression)

79e08da4dae9d445734b214f4fa06a.jpg
 
ChrisLyon said:
Visible is so subjective. Shoot something. Look at it on screen. I don't see how doing a bunch of technical tests changes what your eyes see, or think they see.
Click to expand...

If you have a set running at $10K an hour this is not smart. You want to tell an actor "sorry, the compression was too high?" Or futz with settings while the client waits? Don't test on set. Test off set. Also, often you don't see problems until you are in post making secondary grades, lifting the blacks, etc. . Definitely not a pro approach. Test. Test. Then test again.
 
Les Dittert said:
The compression Red is using in R3D's does not get effected by moving objects, it does not use motion vectors like mpeg or h264.
So a tree moving in the wind is no harder for it than a static tree.
Lots of random detail is stressful. Gamma up the image and look at things like tree bark.
Click to expand...

It does seem to have an easier time with locked tripod and static objects though - could be my imagination...
 
You must log in or register to reply here.
Back
Top