[Warner306]

Dynamic Clipping Explained

The aim of dynamic clipping is to decompress areas of the image that have very few bright pixels concentrated in a very small area. These small patches of bright pixels cause the tone mapping curve to aggressively compress the image to preserve the contrast between the brightest and darkest portions of the image. Frequent compression of small specular highlights can be less optimal for displays with a low peak brightness, as this compression creates lower APLs (Average Picture Levels) that favor the preservation of contrast over brightness.

By clipping a very small percentage of these image areas, the target scene peak for tone mapping can be lowered enough to reduce tone mapping compression and restore a visible amount of brightness to previously compressed portions of the image without completely destroying the specular highlights.

*Dynamic clipping was originally developed by Soulnight at AVS Forum in conjunction with the release of the MadmeasureHDR Dynamic Optimizer tool designed to work with madMeasureHDR measurements files.

Summary of Dynamic Clipping:

• Never clips more than 0-4% of the total pixels in a frame;
  
• Never clips more than 50% of the pixels above 100 nits;
  
• Never clips more than 50% of the pixels corresponding to the 10% brightest pixels in the frame;
  
• The point where highlight clipping is started is never lower than the average nits of all highlights in the frame;
  
• No clipping is performed if there are no areas in the image where the clipped tone mapping target peak would be low enough to produce a visible benefit.
  

Scene-based dynamic clipping:

dynamic clipping targets are calculated for each detected scene change. All frames within a scene are given a compressed target scene peak where clipping of the specular highlights is deemed most beneficial and the frame in the scene with the highest calculated target peak becomes the new scene peak for tone mapping. This typically results in a small of percentage of all frames being clipped to this target peak for any frame peaks that are above the calculated clipped scene peak. Partial clipping of the highlights begins at a highlights knee, which is a point on the PQ curve where big increases in brightness are detected with little pixels added.

Example:

Scene Peak Before Clipping: 999 nits
Frame Peak Before Clipping: 664 nits

Scene Peak After Dynamic Clipping: 468 nits (new clipped target scene peak)
Frame Peak After Dynamic Clipping: 634 nits (after a small % of the brightest pixels are clipped to the scene peak)

When combined with the dynamic target nits, lowering the target scene peak with dynamic clipping can in some cases also cause changes to the display target nits. If the new calculated scene peak is reduced to a value below the min target / real display peak nits, the display becomes capable of representing the scene 1:1 and the target nits is lowered to match the real display peak nits.

Pros & Cons of Dynamic Clipping:

Pros:

• Slight overall increase to average brightness or APL;
  
• More dynamic image with greater color saturation in some scenes.
  

Cons:

• Small loss of specular highlight detail;
  
• Visible reduction in HDR contrast in some scenes;
  
• Possible uncontrolled clipping and blown out highlight detail in some cases.
  

The shape of the tone mapping curve changes based on the tone mapping target scene peak. A more aggressive roll-off is applied to the original scene peak to represent the full dynamic range of the scene.

Original Scene Peak & Tone Curve Shape:


When the tone mapping target is lowered by dynamic clipping, the tone mapping curve straightens due to having less dynamic range to compress. This clips more specular highlight detail, but also reduces overall image compression, thereby creating a slight boost in brightness in the previously compressed portions of the curve.

Clipped Scene Peak & Resulting Tone Curve Shape:


Examples of Dynamic Clipping

*All images courtesy of Soulnight at AVS Forum.

Passengers 4K UHD Blu-ray 

Before dynamic clipping:
frame/avg/scene/movie 302/255/312/527, tone map 296 nits


After dynamic clipping:
frame/avg/scene/movie 290/231/233/527, tone map 214 nits


No tone mapping — simple clipping:
min target / real display peak nits: 150 nits


Passengers 4K UHD Blu-ray 

Before dynamic clipping:
frame/avg/scene/movie 254/255/254/527, tone map 254 nits


After dynamic clipping:
frame/avg/scene/movie 247/231/178/527, tone map 178 nits


No tone mapping — simple clipping:
min target / real display peak nits: 150 nits


Passengers 4K UHD Blu-ray 

Before dynamic clipping:
frame/avg/scene/movie 263/255/263/527, tone map 263 nits


After dynamic clipping:
frame/avg/scene/movie 243/231/135/527, tone map 135 nits


No tone mapping — simple clipping:
min target / real display peak nits: 150 nits

[The Soultaker]

I read your guide Warner306, excellent guide, I was wondering about NGU Standart and NGU Sharp :

"NGU Standard
Similar sharpness to NGU Sharp, but a bit blurrier and less detailed;

Best choice for large upscales applied to lower-quality sources to reduce the plastic look caused by NGU Sharp.

NGU Sharp
Sharpest upscaler and most detailed, but can create a plastic look with lower-quality sources and very large upscales;

Best choice for high-quality sources with clean lines"

As I understood NGU Sharp is best for high quality sources - meaning full blu-rays or good 1080p encodes? I'm not sure what "clean lines" mean? As for NGU Standart is it good for good 1080p encodes, or you recommend NGU Sharp?

Regards.

[Warner306]

I would use NGU Standard with SD content and NGU Sharp with Blu-ray content. NGU Standard is mostly only a better choice when the scaling factor is 3x or larger. Clean lines means the source is not badly compressed and has no aliasing.

[The Soultaker]

I would use NGU Standard with SD content and NGU Sharp with Blu-ray content. NGU Standard is mostly only a better choice when the scaling factor is 3x or larger. Clean lines means the source is not badly compressed and has no aliasing.

I see, my scaling factor is around 1.3x, so I'll stay with NGU Sharp. Thanks for the reply.

[Warner306]

At 1.3x, upscaling won't do a whole lot to improve image quality, but it can help.

[The Soultaker]

At 1.3x, upscaling won't do a whole lot to improve image quality, but it can help.

I guess you are right, but it is a good option.

[The Soultaker]

At 1.3x, upscaling won't do a whole lot to improve image quality, but it can help.

I guess you are right, but it is a good option. But AFAIK image doubling is better than image upscaling, right. The other choice is to use Jinc for image upscaling... 

[gotham_x]

I have no idea if it's really my problem or if it should work like this ,  as an HDR setting madVR , I selected TONE MAP HDR USING PIXEL SHADERS , when I see an HDR movie , the HDR logo is displayed on my OLED panel , I don't think it's normal , or it must be so   The HDR logo should be displayed on the panel for a few seconds with an HDR setting passthrough HDR to display ?  but not with an HDR setting TONE MAP HDR USING PIXEL SHADERS.This is what I've always known , quite confused .

[Warner306]

If you select tone map HDR using pixel shader with output video in HDR format checked, then the output will be an HDR signal. So that can be correct. In this case, madVR is only tone mapping the highlights above the target peak nits that you enter in the control panel. Those few highlights become compressed back into the display rang. The rest of the movie would remain the same.

[gotham_x]

I would like to be sure I understood. I'll explain ,  I have recently installed a second Windows 10 pro operating system on a second M2 SSD , on this new system and with the same settings as madVR and HDR , tone map HDR using pixel shader of the first Windows 10 system  , the HDR logo is displayed on the panel , where instead, as I repeat , on the old Windows 10 operating system housed on the first disk the HDR logo is NOT displayed on the OLED panel , where the contradiction lies ? I ask myself this question , In my opinion there is something wrong and I would like to understand.

[Warner306]

Did you check output video in HDR format under pixel shaders on the new system? This will output with HDR passthrough.

[gotham_x]

Right observation. indeed output video in HDR format was activated , I did not check well , anyway , TONE MAP HDR USING PIXEL SHADERS , work the same with output video in HDR format is activated , where the difference lies ?

[Warner306]

One output in SDR: The entire source range is converted to SDR. Good for a projector, less great for a true HDR display.

Output video in HDR format: Specular highlights too bright for the display are compressed. Better for bright HDR displays and the entire source is not compressed, only the highlights. The output EOTF is PQ SMPTE ST2084 (HDR10) rather than SDR Gamma (2.20 or 2.40).

[gotham_x]

Thank you very much for the explanation , I think I learned something new , however, at the same time I am a bit confused . I read on the AVS Forum forum, which TONE MAP HDR USING PIXEL SHADERS and the tool of Soulnight , MadmeasureHDR Optimizer they are tested and used by beta testers that use an OLED .

I suppose you're against it ?

[Warner306]

Some like it with SDR output and other prefer the display's HDR mode. HDR mode will be brighter with better EOTF tracking. It really depends on personal taste. Most tone mapping development has been focused on projector users, who get a brighter picture with SDR output over HDR output. So most development has been focused on HDR to SDR for projector users.