Welcome to our community

Be a part of something great, join today!

Register 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 :)

Compression Wedges

Gavin Greenwalt

Gavin Greenwalt

Well-known member
Joined
Dec 29, 2006
Messages
7,614
Reaction score
0
Points
0
Age
39
Table of Contents



I did a bazillion compression wedges. Not terribly interesting to look at but the data should be useful.

Process was to take every resolution and reformat to 6k to maintain scale. All framing was equal. Focus I think was dead on. I did it by eye with focus and edge enhance modes for focus. Lens wasn't the sharpest lens in the world: Canon 24-70 L II and focal range was about 45-65. DXO says that all of those focal lengths should have similar lp/mm. Left half is center of frame and focal point. Right half of frame is edge of frame and a little out of focus to highlight flatter less detailed image regions.

It's also very revealing to look at these in Hue Saturation and Value modes. REDCode definitely hits Hue and Saturation much more aggressively than it does luminance. As a result the noise pattern changes dramatically from 2:1 uncompressed to even 8:1 uncompressed. If you want fine noise you really want to shoot at nearly uncompressed.




















 
I was getting ready to do something just like this. You just saved me a ton o time. Thanks!
 
5:1 compression still seems like the sweet spot.

Of course, 3:1 is better but to achieve fluidity on set (i.e. not change mags every few minutes) 5:1 is perfectly acceptable.
 
Thanks for sharing your findings, Gavin.

Gavin Greenwalt said:
It's also very revealing to look at these in Hue Saturation and Value modes. REDCode definitely hits Hue and Saturation much more aggressively than it does luminance. As a result the noise pattern changes dramatically from 2:1 uncompressed to even 8:1 uncompressed. If you want fine noise you really want to shoot at nearly uncompressed.
Click to expand...

I propose a less uniform subject if you plan to conduct more tests in the future. This compression is more visible with greater variations within a whole frame, in shapes and tonalities. This is why this example makes the compression more obvious in a flat part with minute random noise and less on a monochromatic uniform pattern. For example, wide shot with deep focus of a garden with various plants would be more stressing for the codec.
 
Hrvoje Simic said:
I propose a less uniform subject if you plan to conduct more tests in the future. This compression is more visible with greater variations within a whole frame, in shapes and tonalities. This is why this example makes the compression more obvious in a flat part with minute random noise and less on a monochromatic uniform pattern. For example, wide shot with deep focus of a garden with various plants would be more stressing for the codec.
Click to expand...

I agree. The hue/sat testing was incidental so the test is pretty poor conceived as a color compression test, my testing goals were to evaluate resolution vs compression to see if it made more sense to drop resolution and lower compression over highly compressed 6k. I would say the 6k is still 'sharper' than 4k at equal MB/s. The rug was the closest thing I could find to pure random noise and the wall next to it gave me a nice 'flat' area for it to compress a low-detail resolution to see how it handled subtle texture vs hard texture.
 
Gavin,

Thanks for spending the time to do some valuable testing. Can I ask at what crop or magnification are the images that you posted?
 
Bob Gundu said:
A test with movement may also tell a different story too.
Click to expand...

...as will a second of video shot at 12 fps. I did my own testing and found that looking at still images did not give me nearly the information available from looking at a looped playback of video. In my tests, available via this dropbox link

https://www.dropbox.com/sh/bhpx7hx6xavyfae/AAB6IY0Znf5qQ7TEN0HaSIxja?dl=0

the complex pattern of the floor is perfectly stable at lower REDCODE compression ratios, but becomes very distracting as compression rates go up. Pixel-peeping at a single image doesn't tell you how much motion noise you are going to see.
 
Great to see, I'm not sure how transferable this test is to other subjects where the camera is moving for example.

Does anyone have ideas for how such a test could be done?

One test I have been thinking is filming trees in motion with different compression. Lots of detail and contrast.
I guess one would ideally do some moving tests as well. Perfect would of course be if the movement is identical so one can compare them side to side.
I also wondered about the best way to analyse them. I found that Neat Video really helped in analysing compression artefacts by going into the different color modes. There you can see macro blocking and loss of detail very well.

Any thoughts are welcome!
/Andreas
 
More wedges. This is the test I wanted to shoot yesterday as well but never got any sun. As a sharpness test it's useless. The lens just isn't very sharp at that focal length and even if it were the atmospheric distortion between me and trees was visible. Also it's super windy so lots of motion and even the fast shutter I don't think eliminated all the motion blur. But all of them were shot under similar conditions so it's useful to see how they compare.

It's pretty obvious when looking at HSV that compression murders hue and saturation while preserving the luminance until last. So REDCode is a lot like going from 4:4:4 to 4:2:2 to 4:2:0. This is very bad for keying and I would really watch your compression at the cost of resolution for greenscreen work. If it's between 4k and no compression or 6k and high compression pick 4k IMO. You'll end up with very large splotchy chroma choise which will give you a lot of headaches.





 
You must log in or register to reply here.
Back
Top