Difference between revisions of "UM:Dvo Effects Dvo SteadyII"

From DigitalVision
Jump to: navigation, search
Line 166: Line 166:
{{M-Alert|''Density'' will affect the processing speed, so a ''Low'' setting can be used for low complexity scenes to improve speed.}}
{{M-Alert|''Density'' will affect the processing speed, so a ''Low'' setting can be used for low complexity scenes to improve speed.}}
THis parameter lets you optimise based on the resolution of the source media compared to the selected project ''Output Format''. If the source media is
lower resolution than the output format you can use the ''Half'' setting to speed processing. Lower image resolution might be due to
''soft'' source footage, very noisy sources or high resolution scans (e.g scans of 8mm film at 4K or 8K).
Available settings :
{| width="20%" cellpadding="5" border="1" style="border-collapse: collapse;"
|- style="vertical-align:top;"
| Default
|- style="vertical-align:top;"
{{M-Note|Using ''Half'' can reduce processing time by 300%.}}

Revision as of 10:57, 2 April 2019

Darrow-up.png     Previous Next

DVO Steady II
From Release : 2013.2.076


DVO Steady II is used to reduce unsteadiness due to film weave during scanning or camera shakes during capture of footage. The use of advanced motion estimation provides better separation of random instability and pan/tilt movements. DVO Steady II can be used for both film and progressive video material.

DVO Steady II is an updated version of the original DVO Steady tool. New features include splice jump fixing and extension to being a single-pass tool. The new version also includes smaller updates for better results, more options and extra automation.


The tool is organised into three separate pages :

Analyzer Settings regarding the focus of interest of the stabilization and particulars of the material.
Processor Processing parameters are applied to the pictures in accordance with the result from the analysis. These settings related to the amount of stabilization to be performed for achieving a certain smoothness of the result.
Output Settings for the final output rendering, enabling the user to hide the visibility which stabilization has on frame borders.


Every page has a separate Enable button (DEFAULT : Enabled).



This parameter sets the general mode of detection. As well as the the Full image, you can choose to concentrate on other specific regions of interest.

The different modes are :

Full Regards all parts of the image equally.


Centre-Weighted Has a higher focus at the center of the image.
Border-Weighted Has a higher focus at the borders of the image.
Film Frame Only uses the film frame borders in the image. Optionally, this can further be adjusted by specifying which actual borders to include in the analysis. The analysis is separated such that left and right borders are used for horizontal correction, and top and bottom borders for vertical correction.

See below.

Film Corners Only uses the film frame corners in the image during analysis. Similar to the Film Frame mode, the selection of borders also applies here.

See below.

The most tried and tested modes are : Full, Centre-Weighted and Border-Weighted

These will try to stabilize the image content whether unsteadiness is a result of a scanning procedure, damaged sources or just camera movement.

Normal Analyzer Modes

Full, Centre-Weighted and Border-Weighted will be refered to as the normal analyzer modes.

A Note on Film Frame and Film Corner Modes

The Film Frame and Film Corner modes are only intended to be used when the images are overscanned (i.e. the image includes film frame borders and maybe also sprocket holes). The selection of borders should correspond to which sides have overscan available.

These modes will try to only use the frame for stabilization (i.e. will not attempt to correct for camera movement).

Typically, the region of interest should be set manually to where the frame border is, or possibly an edge of any available sprocket holes. An automatic region of interest detection can also be used under some circumstances (see below). There will also be special considerations regarding border selection for these modes.

These modes are fast but should be considered less robust, and the frames themselves may sometimes actually not be the best thing to use for stabilization.



This parameter is used to describe the expected motion content of the material.

Available settings :

Normal The default.
Fast Use Fast mode if you have fast moving material which seems to create breaks in the smoothness of the stabilization using "Normal".

This parameter only has an effect when “Size” is set to “Auto”.



The optimal size parameter is related to several things such as :

  • image size
  • analyzer mode and
  • motion content

Available settings :

Auto The default. The application will determine an optimal setting for Size.
Small Tuning as small.
Medium A larger image, higher motion and more complex analysis mode might require larger sizes.
Large "
X-Large "

Size affects processing time. Larger sizes require more processing and take longer.



Density determines how detailed the analysis will be. The setting is related to the scene motion complexity (i.e. different objects with different and/or complex motion patterns).

Available settings :

Medium This is the Default

Medium is the normal setting. For scenes with higher complexity, try High.

Density will affect the processing speed, so a Low setting can be used for low complexity scenes to improve speed.



THis parameter lets you optimise based on the resolution of the source media compared to the selected project Output Format. If the source media is lower resolution than the output format you can use the Half setting to speed processing. Lower image resolution might be due to soft source footage, very noisy sources or high resolution scans (e.g scans of 8mm film at 4K or 8K).

Available settings :

Full Default

Using Half can reduce processing time by 300%.



This describes the data format of the sources. The analysis will adjust dependent on whether we have Linear or Log input material.

Available settings :

Linear This is the Default

Most of the time Linear will work just as well for Linear and Log formatted images.
However, do not use the Log setting for a Linear source.

Region of Interest


In the normal analyzer modes (see above), the region of interest is simply used to isolate the part of the image which should be analyzed.

If not used to target some specific part, it is usually best to set the ROI to exclude any “black” borders (whether blanking or overscan).

In the Film Frame or Film Corner analyzer modes it is imperative that the ROI is set to where the film frame is aligned (or possibly aligned) with the border of any sprocket holes.


Available settings :

Disable No ROI is used (or rather ROI corresponds to full image).
Specific ROI is set manually using the GUI control overlay or using the individual parameters Left, Top, Right, and Bottom for the offset from respective image edge.


Auto Will find the relevant frame of content automatically removing for instance blanking or overscan. The automatic mode should work most of the time if the material has clean and undamaged borders. In the event of sprocket holes, if they are only available at the far top and far bottom they will be excluded. Otherwise, if running more towards the middle, the Auto ROI will be set to include them.

For Auto ROI, the GUI and Left/Top/Right/Bottom parameters will not reflect the result.

ROI Size

This is the number of pixels from each edge where the frame (active area) starts.

Left / Right : Range : 0 - Width
Top / Bottom : Range : 0 - Height


Default : All Enabled

In the Film Frame/Corner modes the user needs to select which borders can and should be used for the analysis.

Only genuine and completely visible film frame borders can be used, any others must be deselected. The actual film frame is not always completely consistent in size between frames and if this seems to be the case (e.g. stabilization seems ineffective), one must resort to selecting only one of either Left or Right and one of Top or Bottom.

Film Frame mode Left and Right borders will only be used for horizontal analysis and top and bottom only for vertical analysis. As an example, deselecting both Left and Right means no horizontal stabilization will be performed.
Film Corner mode A deselected border implies that both corners on that border are unfit for use in the analysis.



Processing parameters are applied to the pictures in accordance with the result from the analysis.



Range : 1 - 10
Default : 5

Aggressiveness determines the level of smoothness that the stabilization can achieve. The setting should relate to both the amplitude and the temporal naturalness of the unsteadiness which is targeted for stabilization. This parameter can be set separately for horizontal (X) and vertical (Y) stabilization.

Low values are used for fine errors (small and quick e.g. introduced by a scanner/telecine) and does not try to correct for slower and larger movements.

High values take care of fine errors and also handle coarse errors (large and slow) e.g. very old material or handheld/shaky cameras.



The stabilization process is adaptive such that when difficult situations arise it can react and proceed with a robust behavior.

The stabilization process is adaptive, such that the process adapts to maintain correct behaviour. In some situations considered "unsafe", the stabilization effect will be limited to reduce errors.

There are two aspects of safety which are considered : motion and quality.

Range : 0 - 10
Default : 5

High motion frames will be considered less safe (as analysis can become more complex) and more prone to potential “errors”.

Range : 0 - 10
Default : 3

A low quality can indicate a complex scene or other difficulties which will be considered unsafe.

The user does not have to consider this behavior specifically, but rather just set a desired level of safety applied for each aspect. A higher value always corresponds to higher safety. Obviously, if safety is set too high the stabilization will be less effective.



Used to set how much the unsteadiness should be decreased, with separate controls for horizontal (X) and vertical (Y) reduction.

Range : 0 - 100
Default : 100%

Reduction is typically set to 100% to achieve maximum correction.

A bit of residual unsteadiness might be necessary to create a more natural result sometimes e.g. a camera shaking during frame capture resulting in motion blurs.

Max Adjust


Sets a maximum limit (in pixels) for how much the process is allowed to adjust the stabilized frames, with separate horizontal (X) and vertical (Y) controls.

Default : Disabled
Range : 0 - 100
Default : 20 pixels
Splice Fix


Splice Fixing refers to being able to correct for large frame jumps which can occur around scene cuts or what are physical splices in the original reel. This feature is based on finding the number of frames affected on either side of the scene, which can be done automatically (or manually, if necessary).

Values : Disabled, Manual, Auto
Default : Auto
Range : 0 - 25
Default : 7

When set manually, use the setting as a frame count number to the first frame which is considered unaffected (counting from the first / last frame in the scene respectively) e.g. 0 corresponds to disable and 1 corresponds to a correction for the first frame only, and so on.

When a splice fix is actively processing, this feature overrides the effect of Scene Cut Transition (below). “Actively processing” means that the Auto mode has detected an actual jump or that the Frames setting is non-zero in Manual mode.

Scene Cut Transition


Scene Cut Transition refers to special processing modes at the start and end of the scene that takes into account unbalanced temporal information.

This basically means some form of information extrapolation is necessary, and the options are best described according to how safe they are. In the end the different safety settings can be seen as corresponding to how large a correction is allowed to be. For better or worse a Low Safety allows a larger correction than High Safety. The actual behavior may be very similar between modes many times.

Start and End modes are set separately.

Values : Locked, High Safety, Medium Safety, Low Safety
Default : Medium Safety

There is a further option to set the stabilization such that the start/end frame will appear as locked to the original position. The lock option can be useful when a scene is edited and will allow the edit to hold together – such scene edits may be used when one want to change the parameters within a scene for different behaviors.

This functionality is disabled when Splice Fixing is actively processing (see above).



Settings for the final output rendering, enabling the user to hide the visibility which stabilization has on frame borders.

Max Adjust

See Processor section.



The options regarding the output are all connected to the masking of the frame/image border. The most common mode is to use the complete (or interior) frame for stabilization. While the interior will become steady the image/frame borders may instead move around. The masking operation can then be used to blank these potentially moving border areas.

The mask can be set up specifically for each border or can be set to follow the max adjustment set for the processor.

Values : Disable, Follow Max Adjust, Specific
Default : Disable
Left / Right : 0 - Width
Top / Bottom : 0 - Height

If the Processor Max Adjust is disabled, the actual masking will be considered as 0 pixels but the other options in output will still be taken into account (unlike when Mask is set to Disable). The complete masking operation will further be able to include the original blanking area of the source (see below).

When Mask is set to Disable, Blanking and Zoom options (below) have no actual effect.

Soft Border


Creates a soft border at the picture masking.

Default : Disabled

A soft border can give a more natural look.



Blanking is used to describe the original blanking of the source or how it would be set in the source. This information will then be taken into account by the masking operation.

Values : Disable, Specific, Auto
Default : Auto
Left / Right : 0 - Width
Top / Bottom : 0 - Height

See the description of the Analyzer Region of Interest : the setup and behavior is essentially the same except there is no GUI overlay to adjust the blanking manually (the GUI overlay always shows the manually set Regions of Interest of the Analyzer).

If Mask is set to Disable, Blanking has no effect.

If Mask effectively is 0 pixels, then masking will be performed exactly as Blanking is setup.



The zoom option allows the final masked result to be zoomed.

Values : Disable, To Blanking, Zoom Letter/Pillar Boxed, Zoom Fill Full Screen
Default : Disable

Settings :

Disable No zoom
To Blanking Zooms to the blanking setup (for the source)
Zoom Letter/Pillar Boxed The masked output is centered and zoomed such that it fills either letter or pillar box whichever is reached first. The original pixel aspect ratio is maintained
Zoom Fill Full Screen The masked output is centered and zoomed such that is fills the full screen and if necessary clips image either at top/bottom or left/right depending on what is necessary. The original pixel aspect ratio is maintained

All modes keep the correct aspect ratio.

Darrow-up.png     Previous Next