Standard Picture Style vs Cinestyle

Started by test, May 19, 2012, 02:10:32 PM

Previous topic - Next topic

0 Members and 1 Guest are viewing this topic.

test

OUTLINE

There has been much debate over how useful the Technicolor Cinestyle Picture style is, whether it helps post production, or harms the final image with banding. This page contains the results of a fair test between the Technicolor Cinestyle picture style and the Standard Canon Picture Style. The aim of these tests is to highlight the areas where using the Cinestyle picture style can help or harm your final video frames compared to using a non-flat built in picture style.
These tests DO NOT intend to compare Technicolor Cinestyle with any other flat or "cinema" picture style. The tests reflect a professional workflow from the raw* video files to the final image. BOTH Standard and Technicolour Cinestyle video clips were treated to exactly the same settings, using exactly the same software, captured at exactly the same settings**
*Raw video in this case meaning un-changed from the state they were recorded in from the camera, NOT Raw such as Raw uncompressed photo files.
**The exact same settings except of course for the changing of the picture style.

TEST CONDITIONS

Filming Equiptment used:

Canon 5d MkII
Canon 50mm F1.8 Prime

Firmware:
Magic Lantern Unified xmas Build

Camera Settings:
F2.0 - 1/50 - ISO100
1920x1080 @23.976

Workflow


  • Captured 1920x1080 23.976fps h.264 video from Camera - Both Picture styles are set to 0 Sharp -4 Contrast -2 Saturation 0 Tone
  • Converted to Cineform video files using Cineform NEO HD LINK software - Part of this process involves interpolation from 4:2:0 to 4:2:2 and also the video files are now using 10bit colour space

  • Video clips are brought into Cineform FirstLight for post production - this does not require any re-rendering as it changes meta data in the video files, meaning they are simply re-interpreted - retaining 100% of the quality, this process is also fully non-destructive and can be fully reversed back to the original files with 100% accuracy. This software also allows for interpolation between the shades, helping to prevent banding that may occur. The effect of the changes made in this software is applied in realtime to the clips regardless of what program they are also open in.
    THE ONLY DIFFERENCE BETWEEN THE TWO VIDEO FILES IS THAT THE TECHNICOLOUR CINESTYLE CLIP HAS THE CORRECT LUT APPLIED AT THE END OF THE COLOURING PROCESS, AS ADVISED BY TECHNICOLOR.)

  • While video clips are still loaded in Cineform First Light, they are also imported to Sony Vegas Pro 10.

  • Both clips sit at 3 seconds into the video and the frames are saved off at full resolution/quality as jpeg files, which are the same ones provided on this website.

NOTES

As this test is supposed to demonstrate how well post production can be applied to the images, the tests are examples of they various things one may do the clips. These do not include any sharpening or denoising, only exposure, contrast, saturation and LUT style coloring. The best way to view the results would be to have them in 2 browser tabs and flick between them, as such i have provided links instead of thumbnail images.
I will provide objective notes with each set of results, I have my own preference of course, but I will try to only mention details to look out for if i have noticed them. Whilst different conditions may bring out the worst or best in either of the picture styles and indeed this could cause a bias in favour of one or the other of the picture styles, the only way to eliminat this would be to do another series of tests in a completely different test condition. For example; Flat picture styles have a reputation of suffering slightly when used on already flatly lit scenes, such as overcast external shots. Perhaps In the future I will get round to testing this.

TEST RESULTS

Image is overexposed by a value of 1

STANDARD: www.tommajerski.com/tests/picturestyle/Over%20Exposing/Standard.jpg
CINESTYLE: www.tommajerski.com/tests/picturestyle/Over%20Exposing/Cinestyle.jpg
Notes:

  • The overexposed area of the back wall on the left appears more blown out in the cinestyle shot.

  • On the right hand side of the quill, there is subtle banding that seems to get slightly worse on the cinestyle shot.

  • On the far right of the image, is a dark patch which seems to contain compression artefacts/noise on he standard shot but not on the cinestyle shot.

Image is Underexposed by a value of 0.5

STANDARD: www.tommajerski.com/tests/picturestyle/Under%20Exposing/Standard.jpg
CINESTYLE: www.tommajerski.com/tests/picturestyle/Under%20Exposing/Cinestyle.jpg
Notes:

  • There is very subtle banding on the right hand side of the quill that looks to be only visible in the cinestyle shot.

  • On the far right of the image, is a dark patch which seems to contain compression artefacts/noise on he standard shot but not on the cinestyle shot.

Saturation is increased
STANDARD: www.tommajerski.com/tests/picturestyle/Increasing%20Saturation/Standard.jpg
CINESTYLE: www.tommajerski.com/tests/picturestyle/Increasing%20Saturation/Cinestyle.jpg
Notes:

  • Same as before, more banding on Cinestyle, more compression artefacts on Standard.

  • Not visible in these results, but when pushed slightly further, nasty blotches would appear on both images, but they occured sooner on the standard profile.

Contrast is increased
STANDARD: www.tommajerski.com/tests/picturestyle/Increasing%20Contrast/Standard.jpg
CINESTYLE: www.tommajerski.com/tests/picturestyle/Increasing%20Contrast/Cinestyle.jpg
Notes:

  • Not too much difference here, hints and banding on cinestyle, nothing too serious but the shadow noise from the standard seems to have been crushed away.

  • There does seem to be a slight red shift on the standard picture style despite white balance remaining constant throughout testing, although it could be that that is indeed more faithful than the cinetsyle image, rather than it being a colour bias.

Contrast is decreased
STANDARD: www.tommajerski.com/tests/picturestyle/Decreasing%20Contrast/Standard.jpg
CINESTYLE: www.tommajerski.com/tests/picturestyle/Decreasing%20Contrast/Cinestyle.jpg
Notes:

  • This test was bound to look better on the cinestyle profile as reducing contrast is closer to the original flat look as captured.

  • Strange blotches on the Standard image, visible especially on the computer keyboard on the left side of the image.

Heavy Colour correction applied
STANDARD: www.tommajerski.com/tests/picturestyle/Extreme%20Colourising/Standard.jpg
CINESTYLE: www.tommajerski.com/tests/picturestyle/Extreme%20Colourising/Cinestyle.jpg
Notes:


  • Both images exibit slight banding on gradiated areas, slightly worse on the cinestyle shot.

  • Tones seem generally more affected on the cinestyle profile than they are on the Standard one - not necessarily good or bad, depends on your viewpoint.

  • Again, slightly more detail visible in darker areas, such as on the computer keyboard on the left.

OPINION

WARNING:This section will contain opinion, not fact - opinion based on these results and experience using cinestyle for making films. I am in no way connected to any camera company or post production company. For these reasons, this section can not possible contain anything that is "wrong" or "incorrect". You may disagree, or agree, but neither of us can be wrong when its just an opinion.
The two most common things that stand out for me in these tests are that Cinestyle produces slightly more banding and that Standard Produces more noise/artefacts in the shadows.

Which would I consider worse? Well the Banding occurred on pretty much all of the shots, and the noise/artefacts only occured on some of the shots. Does this mean that I prefer the Standard picture style?
No, i still prefer cinestyle. Why? Well, for my tastes the banding was never too bad, perhaps it may be worse in other conditions, more tests under the same conditions would prove or dispove this - but the artefacts in the shadows for my tastes are unacceptable. It's true that when the contrast was increased, the artefacts in the standard image pretty much went away, but I prefer a less contrasty image, and i think that for my style of working, i would encounter these artefacts too often.

Surpise surprise, it ultimatly depends on how you want your final images to turn out. If you like the crushed contrasty look - Standard would probably be great for you, and produce best results. But if youre either 1:Not sure when at the filming process or 2: You dont like crushed backs - A flatter picture style, or indeed the technicolor Picture style would probably be your best bet. Perhaps there are better "flat" picture styles out there, i have not evidence online to show this, but i find the Technicolor cinestyle to be the best flat profile - maybe my next set of tests should show why.

I have often seen other tests online that demonstrate far worse banding, but I have always noticed that they are either using the raw h.264 files without using a decent intermidiate codec and or they are not editing in the full colour space. Generally, professionals would never edit with h.264 files, they would use a proper intra-frame codec such as Cineform or ProRes and would always edit in a 32 bit colour space.

Many people would argue that they do not see the point in shooting flat because it requires extra post production  - in such cases where you do not want to do any post production, there is no point in using a flat picture style, for my own work, i always do some post production, even if its very slight. When using a real time colour corrector, its hardly time consuming to do so.

Perhaps the best way to think of it is this:

At the moment, we are stuck with a crappy 8bit compressed video format, so the image is never going to be perfect. So you can either A) Shove the crap into the shadows or B) spread it out across the range. Now these are not technically accurate statements, but i think its a fair way of weighing up the pros and cons of each option. If you want A - shoot standard, if you want B - shoot cinestyle/flat.

1%

Look for anything other than cinestyle. There is Flaaat and the other one... at the minimum just go to neutral, turn off sharpening and drop contrast + saturation a little.

Cinestyle is more or less for inter cutting with film.  Using it regularly will just give you headaches and cause banding + other problems. I actually had less latitude in post before the footage broke up trying to squeeze out more dynamic range with these picture styles. The new consensus is that they make things worse.

If you're getting good results its because you're converting to cineform which expands the color space to 4:2:2/4:4:4 upon conversion. ProRes or DNxHD will work too.... as stated.


http://www.shutterangle.com/2012/canon-picture-styles-shooting-flat-or-not/

aace

If I may add to this. To help with the banding issues you could use the technique in the video below. I've invested both in Cineform and Neatvideo and together they transform a very compressed image into footage ready for moderate color grading. Link to the tutorial here:


bart

Interesting video and great results. Add noise does debanding and dithering is not a very obvious use. But good to know.

DFM

@aace : If you're working in After Effects* there is absolutely no point in upsampling the footage before import. Switching AE to 16 or 32-bit working space means everything you do on the timeline to clips with a lower bit depth is automatically interpolated if the effect supports that bit depth. Similarly if you export a 16 or 32-bit timeline to an 8-bit media file, After Effects automatically adds a dither pass to mitigate banding. You won't normally see it happen but you can force dithering to affect the preview buffer by applying a null 8-bit effect (e.g. Arithmetic with everything set to zero) on an adjustment layer.

*Converting to a mezzanine format (e.g. ProRes) is useful for NLEs where you need to scrub the timeline a lot, as it's less demanding on your CPU to rapidly-decode the frames. After Effects doesn't work like that, it has an internal buffer with format-agnostic pixel data for every frame (users see it as the RAM Preview).

You shouldn't export to RGBA unless the footage contains a true alpha channel (e.g. footage assembled from a PNGA sequence or a timeline with open masks). It bloats the file with null data, which makes it harder for subsequent software to process.

deleted.account

Sound like someone with a technical knowledge of AE. Excellent. :-) A query or two.

Isn't it rather academic though because unless using Dynamic Linking and not everyone does, the input into AE with regard to grading after an edit won't be the original MOVs off camera anyway but an intermediate format of some description?

Further, that intermediate format unless RGB output will almost certainly be limited range. Canon chose to encode using full luma and chroma ie: JFIF as many DSLR camera manufacturers do in order to make the most of the data. Also flagging the h264 stream to signal 'fullrange' data, but typical output from an encoder receiving full range is to scale it into 16 - 235.

I'd also suggest one reason to transcode or convert to EXR or similar is if importing MOV's directly into a 32bit linearised AE project is that it appears impossible to get a correct degamma'd result because AE guesses sRGB iec61966 for the MOV's, so to linearize,  AE appears to apply a reverse sRGB 2.2 gamma which is incorrect on a rec709 source.

Tell AE to interpret footage as rec709 and it assumes 16 - 235 as range on a full range source which again results in an incorrect result, shown by a combed histogram.

So removing the full range flag and import remuxed mp4 into AE and mediacore imposes rec709 as format, correct but again uses 16-235 range as input on a full range source, so there appears no route other than to feed RGB perhaps as OpenEXR or a range limited YCC intermediate transcode to produce correct results for a linearised 32bit workflow?

heanz

Please allow me resurrecting this thread, because it's the first google result when I search about compression artefacts in DSLR videos. So I wanted to add something I figured out.

The reason, CineForm's FirstLight has artefacts on 8-bit material, and not on AFter Effects is not your dithering and interpolating. FirstLight changes the metadata of the file and not the actual footage to "color correct" it. So there is no 32-bit floating. And Premire / After Effects do the upscaling internally, as DFM stated.

CineForm revealed to be completely useless when trying to avoid artefacts in color grading when using a good NLE with 32-bit-float.