Welcome, Guest
Username Password: Remember me

TOPIC: orange filter B/W effect

Re: orange filter B/W effect 9 months, 2 weeks ago #215893

  • jwrl
  • Moderator
    Pro User
  • OFFLINE
  • Moderator
  • Posts: 13419
  • 9 months, 2 weeks ago
I have a version of the HSP profile somewhere, based on a paper that I had found I think at an American University, but Finley's work looks very similar. It's obviously the original.

However in this context it's not really appropriate. The colour profile of panchromatic film isn't like the HSL, HSV or HSP. It's lumpier, and has much hotter blues than any of those. Perhaps the nearest is somewhere between the V component of HSV and the L component of HSL.

I have developed my profile empirically, having done a lot of experimentation with greyscale curves. I make no promises as to accuracy, and in fact couldn't anyway. The actual values would vary depending on the panchro film stock that you were trying to emulate. Mine is just a generic panchro feel. If I was being a purist I would linearise the RGB as Finley does, then convert it to luma, then convert that luminance image back to RGB.

At the moment the effect works and I'm not sure that I want to recalculate the profiles for a linearised version.

Re: orange filter B/W effect 9 months, 2 weeks ago #215905

  • hugly
  • OFFLINE
  • Platinum Boarder
  • Posts: 27533
  • 9 months, 2 weeks ago
After some further tests, with orange and red filter emulations, banding sets quite narrow limits for acceptable results on blue skies, even with 10-bit 4:2:2 media and export.

That doesn't even come close to what can be achieved on film.
It's better to travel well than to arrive...

Re: orange filter B/W effect 9 months, 2 weeks ago #215907

  • schrauber
  • Pro User
  • OFFLINE
  • Platinum Boarder
  • Posts: 4904
  • 9 months, 2 weeks ago
jwrl wrote:
.. HSP profile [..] However in this context it's not really appropriate..

That's how I see it, too.

Some time ago I was interested in HSP because ..
:pinch: Warning: Spoiler!
Mainly automatically translated
--------------------------------------------
Software: Lightworks 2020.1.1 & Beta 2021.x; || Windows 10 Home 1909, 64 Bit
Hardware: Intel i5-4440 (3,1 GHz); || RAM: 8 GB; || Intel HD Graphics 4600 (can use max. 2 GB of shared RAM)
Last Edit: 9 months, 2 weeks ago by schrauber.

Re: orange filter B/W effect 9 months, 2 weeks ago #215909

  • schrauber
  • Pro User
  • OFFLINE
  • Platinum Boarder
  • Posts: 4904
  • 9 months, 2 weeks ago
hugly wrote:
.. banding ..

Does it help if you increase the color depth of the original with for example Artifacts-blur_20.fx before doing a gray scale? (set at least 16 bit GPU precision)

If that is not enough, maybe at the end of the routing, the UncountourV2.fx effect?
Mainly automatically translated
--------------------------------------------
Software: Lightworks 2020.1.1 & Beta 2021.x; || Windows 10 Home 1909, 64 Bit
Hardware: Intel i5-4440 (3,1 GHz); || RAM: 8 GB; || Intel HD Graphics 4600 (can use max. 2 GB of shared RAM)

Re: orange filter B/W effect 9 months, 2 weeks ago #215913

  • hugly
  • OFFLINE
  • Platinum Boarder
  • Posts: 27533
  • 9 months, 2 weeks ago
I think the general problem with such effects is that they hit the whole image, but are needed only in some selected (read masked) areas.

As long as Lightworks doesn't have the ability to manually draw or paint masks and track them, I believe users are better off when using software which has.
It's better to travel well than to arrive...
Last Edit: 9 months, 2 weeks ago by hugly.

Re: orange filter B/W effect 9 months, 2 weeks ago #215940

  • jwrl
  • Moderator
    Pro User
  • OFFLINE
  • Moderator
  • Posts: 13419
  • 9 months, 2 weeks ago
hugly wrote:
I think the general problem with such effects is that they hit the whole image, but are needed only in some selected (read masked) areas.

That isn't really what's happening here. We are trying to emulate as closely as possible the effect of an optical filter. Masking has nothing to do with it. I suspect that kalimerox's fix of using a LUT with S-log media then applying the effect is the closest that it will be possible to get the classic look. Trying to achieve that look with selective masking will quickly become a nightmare (from bitter experience).

jwrl wrote:
.. If I was being a purist I would linearise the RGB as Finley does, then convert it to luma, then convert that luminance image back to RGB ... However in this context it's not really appropriate.

schrauber wrote:
As I understand it, Finley's formula does all this.
This image is hidden for guests. Please log in or register to see it.

I think within the brackets linearized luma is calculated, and sqrt( luma ) converts to the RGB characteristic.

In my post I was really referring to linearisation in general rather than Finley's approach, which is an empirically derived approximation in any case. A more accurate example might be:

   float3 linear = ((Inp.r < 0.04045) && (Inp.g < 0.04045) && (Inp.b < 0.04045))
                 ? Inp / 12.92 : pow ((Inp + 0.055.xxx) / 1.055, 2.4;

   // Do some stuff here then get out

   Inp.rgb = ((linear.r < 0.0031308) && (linear.g < 0.0031308) && (linear.b < 0.0031308))
           ? linear * 12.92 : (1.055 * pow (linear, 0.4166667)) - 0.055;
   Inp.rgb = saturate (Inp.rgb);

That example is based on a code fragment that user abelmilanes built into his film exposure effect. He actually went further and converted the linearised image to a Cineon profile to get a more film exposure-like characteristic.
Last Edit: 9 months, 2 weeks ago by jwrl.

Re: orange filter B/W effect 9 months, 2 weeks ago #215945

  • RWAV
  • Moderator
    Pro User
  • OFFLINE
  • Moderator
  • Posts: 7068
  • 9 months, 2 weeks ago
Adding and tweaking a Tint to simulate just the right amount of filtration and the Greyscale tool to replace all the developing and printing variables will provide infinite paths to an Ansel-Adams-like Nirvāṇa.

This attachment is hidden for guests. Please log in or register to see it.
BETA System
Microsoft Windows 7 Professional 64BIT
HP Z800 Workstation

Re: orange filter B/W effect 9 months, 2 weeks ago #215949

  • hugly
  • OFFLINE
  • Platinum Boarder
  • Posts: 27533
  • 9 months, 2 weeks ago
jwrl wrote:
hugly wrote:
I think the general problem with such effects is that they hit the whole image, but are needed only in some selected (read masked) areas.

That isn't really what's happening here. We are trying to emulate as closely as possible the effect of an optical filter. Masking has nothing to do with it.

I didn't refer to the attempts to emulate the effects of colour filter on BE film stock, but the use of effects to reduce banding in this context, i.e. this dialogue:

hugly wrote:
After some further tests, with orange and red filter emulations, banding sets quite narrow limits for acceptable results on blue skies, even with 10-bit 4:2:2 media and export.

That doesn't even come close to what can be achieved on film.

schrauber wrote:
hugly wrote:
.. banding ..

Does it help if you increase the color depth of the original with for example Artifacts-blur_20.fx before doing a gray scale? (set at least 16 bit GPU precision)

If that is not enough, maybe at the end of the routing, the UncountourV2.fx effect?

hugly wrote:
I think the general problem with such effects is that they hit the whole image, but are needed only in some selected (read masked) areas.

As long as Lightworks doesn't have the ability to manually draw or paint masks and track them, I believe users are better off when using software which has.


RWAV's experiment, combining Tint with Greyscale, shows the banding clearly up right in the sequence viewer.
It's better to travel well than to arrive...

Re: orange filter B/W effect 9 months, 2 weeks ago #215952

  • RWAV
  • Moderator
    Pro User
  • OFFLINE
  • Moderator
  • Posts: 7068
  • 9 months, 2 weeks ago
I used some crap YouTube AVC source material in a 24 fps project - just an illustration
BETA System
Microsoft Windows 7 Professional 64BIT
HP Z800 Workstation

Re: orange filter B/W effect 9 months, 2 weeks ago #215953

  • hugly
  • OFFLINE
  • Platinum Boarder
  • Posts: 27533
  • 9 months, 2 weeks ago
From what I've seen here so far on blue skys with Rec 709 4:2:2 material, yellow filters work to some extend, orange and red create almost always banding. I think, for reasonable results, without effects to suppress banding, footage captured in higher bit depths (10,12, or even more) is necessary.

Alternatively, people can treat the image with "anti-banding" filters, but I doubt that we will find anything which works without masking against banding on blue skies effectively and leaves the rest of the image untouched.

Edit: And at the very end, it needs something like 10-bit HDR export to show something reasonable on Youtube - a missing feature, by the way.
It's better to travel well than to arrive...
Last Edit: 9 months, 2 weeks ago by hugly.

Re: orange filter B/W effect 9 months, 2 weeks ago #215959

  • schrauber
  • Pro User
  • OFFLINE
  • Platinum Boarder
  • Posts: 4904
  • 9 months, 2 weeks ago
hugly wrote:
I think the general problem with such effects is that they hit the whole image, but are needed only in some selected (read masked) areas..

This is exactly the main feature of these effects.
These effects produce an automatic mask internally which can be fine-tuned with just two sliders. The automatically generated mask can be displayed in red for fine adjustment.
www.lwks.com/index.php?option=com_kunena&func=view&catid=7&id=209274&limit=15&limitstart=15&Itemid=81#209443

Of course such automatic masks are not perfect, but perfection is usually not necessary with this mask.
Red areas of the mask do not always mean that all pixels in the mask are changed by the blur process. Therefore not all mask inaccuracies are visible.
With good original material, the threshold value for mask generation can be set so low that with 8-bit original color depth, only pixels with 1-bit color difference are blurred, so that small mask inaccuracies are probably rarely noticeable?

So far, however, there are not enough reports about these effects. Theoretically I could also add a color selection (e.g. light blue) if mask inaccuracies would otherwise blur other colors. Whether this is necessary, however, would also have to be tested.
Mainly automatically translated
--------------------------------------------
Software: Lightworks 2020.1.1 & Beta 2021.x; || Windows 10 Home 1909, 64 Bit
Hardware: Intel i5-4440 (3,1 GHz); || RAM: 8 GB; || Intel HD Graphics 4600 (can use max. 2 GB of shared RAM)

Re: orange filter B/W effect 9 months, 2 weeks ago #215964

  • hugly
  • OFFLINE
  • Platinum Boarder
  • Posts: 27533
  • 9 months, 2 weeks ago
Ah, ok.

I'm still experimenting with the build-in Greyscale effect with different images in order to see how to create some reasonable results with it. The 'Channel Contrast' sliders don't help me much on this, but 'Channel Weight' does. I believe for further adjustments a CC after the Greyscale can make sense.

Just one not very challenging example (saved frame converted to JPG for file size), the original:

This image is hidden for guests. Please log in or register to see it.

RGB (100,0,0) for a red filter

This image is hidden for guests. Please log in or register to see it.

RGB (<default>, <default>, zero) for a yellow filter

This image is hidden for guests. Please log in or register to see it.

Yellow looks quite good, but red creates banding up right. Do you think you can compensate for that somehow, without affecting the rest visibly? The original attached.
Attachments:
  • Attachment This attachment is hidden for guests. Please log in or register to see it.
It's better to travel well than to arrive...
Last Edit: 9 months, 2 weeks ago by hugly.

Re: orange filter B/W effect 9 months, 2 weeks ago #215971

  • RWAV
  • Moderator
    Pro User
  • OFFLINE
  • Moderator
  • Posts: 7068
  • 9 months, 2 weeks ago
Went on to repeat previous ideas with uncompressed media - there is no banding.

Banding is caused by AVC style block compression - in the original post the source material already had banding - partly camouflaged by the dithered Chroma. Remove the Chroma leave the Luma - the banding is predictably exaggerated

Did a variation - rather than using the LW Colour Tint tool made an oversized uncompressed colour image to key in as the 'filter' - with uncompressed material that too did not result in banding. However it was necessarily complicated - the LW Colour Tint tool is just fine.

The concept of adding a coloured filter over the shot scene - then representing the result as a grey-scale image is parallel to placing a filter over a lens and shooting on grey-scale film stock.

The process is one of achieving a tonally distorted 'look' it is about creative judgement - pretty difficult to express as numbers - a matter of taste and style. The answer is not 42 - the secret to making a great looking shot is not brightness = sqrt( .299 R2 + .587 G2 + .114 B2 )

Be creative, please. Trust in one's own judgement.

Was able to source and load contemporary colour images from Yosemite national park - some closely mimicking classic Ansel Adams compositions.

The combination of a 'digital' colour filter and image processing settings of the LW Colour Tint tool easily allows for results similar to Ansel Adams' tonal treatments.
BETA System
Microsoft Windows 7 Professional 64BIT
HP Z800 Workstation

Re: orange filter B/W effect 9 months, 2 weeks ago #215974

  • hugly
  • OFFLINE
  • Platinum Boarder
  • Posts: 27533
  • 9 months, 2 weeks ago
RWAV wrote:
The combination of a 'digital' colour filter and image processing settings of the LW Colour Tint tool easily allows for results similar to Ansel Adams' tonal treatments.

I think you can live with that I disagree with your opinion.
It's better to travel well than to arrive...
Last Edit: 9 months, 2 weeks ago by hugly.

Re: orange filter B/W effect 9 months, 1 week ago #215985

  • schrauber
  • Pro User
  • OFFLINE
  • Platinum Boarder
  • Posts: 4904
  • 9 months, 1 week ago
hugly wrote:
.. but red creates banding up right. Do you think you can compensate for that somehow, without affecting the rest visibly? The original attached.

Thanks for the examples.
Since I don't have a 4-K monitor, I could hardly see the banding in full screen, but unscaled it was more clearly visible (or with temporary zoom effect in Lightworks at UHD project resolution).

Here a PNG-image that uses the default settings of the effect "Artifacts-blur-Test20". :
This attachment is hidden for guests. Please log in or register to see it.


What does it look like on your monitor?

By the way, in my test it also worked, with 8 bit GPU precision, which indicates that the banding is probably jpg artifacts or something similar.
Attachments:
  • Attachment This image is hidden for guests. Please log in or register to see it.
  • Attachment This image is hidden for guests. Please log in or register to see it.
Mainly automatically translated
--------------------------------------------
Software: Lightworks 2020.1.1 & Beta 2021.x; || Windows 10 Home 1909, 64 Bit
Hardware: Intel i5-4440 (3,1 GHz); || RAM: 8 GB; || Intel HD Graphics 4600 (can use max. 2 GB of shared RAM)
Last Edit: 9 months, 1 week ago by schrauber.
Time to create page: 0.49 seconds
Scroll To Top