Example of an interlaced picture. Note how the action is brokendown into two. The person is moving.
On the video display one field is displayed, then the other athalf a frame time later.
ReelSmartTM FieldsKitTM Users' Manual.
ReelSmart FieldsKit currently consists of 3plugins:
What is ReelSmartTMFieldsKitTM?
ReelSmart FieldsKit is a complete plugin toolkit tomanipulate video fields with higher quality and more flexibilitythan current applications generally allow. We hope this will beuseful to effects people,who generally need full frames and notinterlaced video to do their work, and video editors alike.
ReelSmart Deinterlacer uses adaptive motion techniquesand clever models for deinterlacing. ReelSmart Deinterlacerbuilds full frames of much higher quality from video fields.
ReelSmart Reinterlace and ReelSmart Pulldown allowmuch more flexible interlacing and pulldown removal/addition thanmost applications provide.
ReelSmart FieldsKit is a great companion to other ReelSmartmodules such as Twixtorand Motion Blur.
ReelSmart FieldKit worksin Adobe's After Effects, Apple's Final Cut Pro,Pinnacle's Commotion, Adobe's Premiere 6.0, Discreet'scombustion and other programs that support After Effectscompatible plugins.
What are fields
1) Simply put, an interlaced video frameconsists of 2 fields. One field consists of all even lines of aframe, and the other field consist of odd lines. A video device(monitor, TV, etc) has a beam that traces, or draws, one field,then goes back and draws the other. As such, one field isdisplayed first, then the other is displayed half of a frame timelater.
Example of an interlaced picture. Note how the action is brokendown into two. The person is moving.
On the video display one field is displayed, then the other athalf a frame time later.
2) You have two fields. One field is displayed first in time,and then the other. If the top scan line is part of the fieldthat is displayed first (that is, the field of odd scan lines isdisplayed first) then we say that the video is "Upper FieldFirst." If the second-to-top scan line is part of the fieldthat is displayed first (that is, the field of even scan lines isdisplayed first) then we say that the is "Lower Field First".
The rule of thumb when specifying "field first"for your footage: if you choose lower or upper field firstand the plugin produces material that is overly jittery, or jumpsback and forth, then you probably chose the incorrect option.
Note that you cannot assume all of your source material willhave the same field first-ness.
MORE INFO :
From Trish and Chris Meyer :
http://www.dv.com/magazine/1998/0898/meyer0898.html
Usage Notes:
Important: Before you start, make sure you understand howyour application deals with field display. (For examplethe field display control in Combustion is underWindows->Field Mode menu). When using ReelSmart FieldsKit youhave to make sure that your application doesn't do any processingwith the fields (like make a whole frame out of one field forconvenient display, as is done in After Effects or Combustion).Also, make sure that Field Rendering is off (all hostapplications need to have this turned off!). In essence,FieldsKit needs to take control of fields, not the application.The document
FYI, some interlaced video has an odd number of fields. Inthis case the two fields won't be of the same height. TheReelSmart FieldsKit handles this case by duplicating a lineduring the deinterlacing process.
Note that we will use the symbols < , = , > in the option labelsas amnemonic, to imply that few frames, the same number of frames, ormore frames are created as a result of applying the plugin.
Please refer to the Application Note "
Also note that we will use the term"combs" to refer to areas where there is significantmotion in a frame with fields, because the resulting areas look"combed."
Two areas with "combs" are shown.
Related ReelSmart Modules:
ReelSmart FieldKit is a perfect companion to ReelSmartTwixtor (which asks you to turn your favorite app fieldhandling off):
* In general when the camera does not move you should obtainbetter results with Twixtor by first restoring better full framesusing FieldKit and compensating then appropriately the durationin Twixtor (you will now have 2 times more frames). So in anutshell, we prefer to provide simple self-contained modulesrather then beef up Twixtor so it ends up with zillions ofcontrols you might never need and a price point you cannotafford.
* Because Twixtor needs to take control of fields handling, itforces people to split their rendering in two passes if they wantto add 3:2 pulldown for example. Having the whole field handlingchain as filters as opposed to IO attributes, allow users ofapplications that do not support frame rate conversion (thatmeans for instance going from 25 FPS to 30 FPS) theflexibility to better streamline handling of field-basedmaterial.
ReelSmart Motion Blur can be used as a post-processafter Deinterlacing with FieldKit, to simulate the true effect ofa shutter if you use it simply to go from 30 FPS interlaced to 30FPS progressive. Using ReelSmart Motion Blur can sometimes alsohelp deal with some crawling that persist after defielding yourmovie with FieldKit. Also, since Twixtor requires that you turnFields to Off when running your application, FieldsKitmight be an ideal companion for what you are doing, allowing forinstance to add a 3:2 pulldown or simply reinterlace after havingapplied a bunch of process to your footage.
Deinterlacer
Overview:
The Deinterlacer creates a frame for eachindividual field. In the simple case the plugin throws out everyother scanline, then create new scanlines by interpolating theremaining scanlines.
 Original material |  Every other scanline thrown out |  Thrown out scanlines filled with interpolated pixels. |
"But wait!" yousay. Thetea kettle isn't moving, why are we throwing away every otherscanline there? Rest assured that Deinleracer includes some toolsto mask out where pixels are thrown out, and subsequently, wherepixel interpolation takes place.
The options, described step by step:
1) Select the Source Layer to Deinterlace
Source Layer :
The layer to deinterlace. Disclaimer: A plugin cannot by itselfextend the duration of a clip. It needs help from the user andthe usage of an external layer. Make sure to read the applicationnote on how to extend the duration when applying a RE:VIsionEffects plugin.
Note: In After Effects, if there are any filters applied tothe source footage, then "Source Layer" should be apre-composition, otherwise filters applied to the source layerwill not be visible. Other applications may require you to putyour source in containers in order to see other applied filters.
2) Select the timing mode. Do you want to make a frame fromthe first field of each frame, or do you want to make a framefrome each field of each frame (which creates 2 times as manyframes as the original)?
Timing Mode:
3) What is the field order of your source material?
Field Order:
4) Choose a method to interpolate pixels
Fill Method:
We provide various methods to spatially interpolate/create pixelsin the non-exisiting field scanline when making a full frame forit.
Note: Best 3 Neighbors and Best 5Neighbors tries to give better results on diagonallines.
| Fill Method: None |  Fill Method: Duplicate |  Fill Method: Blend |
| Fill Method: Best 3 Neighbors |  Fill Method: Best 5 Neighbors |
Creating a mask, based on motion, wherepixels are "thrown out" and then interpolated fromnon-thrown out scanlines.
The remaining steps, except for the last two, allow you to setup a region where the interpolation of scanlines is to takeplace. Why would you want to do this? Well, for example, if thecamera is still and only a foreground object is moving then youreally only want to "throw out" scanlines andinterpolate pixels where the object is moving. The following 7options allow you to specify where interpolation should occur.
IMPORTANT: If you specify that you want to use a motiondetection scheme, a frame is created from a field byinterpolating the field's scanlines only where there is motion.Otherwise, the pixels are obtained by the other field ofthe same frame. Using the options that follow, you implicitlycreate a "motion mask" that specifies where pixels areto be interpolated. You can even view this motion mask todetermine if your settings are appropriate.
IMPORTANT: You want to set the "motion mask"to be on everywhere where you see combed behavior in the images(that is, where there is significant motion!). You can alternatethe view mode in order to see where the motion mask is turned on.
View Motion Mask:
This allow you to display the motion mask (areas where tointerpolate scanlines). The areas where pixels will beinterpolated are displayed in Red. The objective is to makesure that you see that all "combs" are covered byred. Make sure you don't leave the Deinterlacer in thismode before you launch a render. HINT: The motion mask created inthe Deinterlacer is stored in the RED channel. You may want touse the red channel as a mask to run alternative processes otherthan those included in the deinterlacer.
The picture examples for the motion settings are shown withthe View Motion Mask option on.
5) Do you want to want to interpolate pixels only wherethere is motion?
Detect Motion:
Note, the choice for Detect Motion isimportant. Using 3 Frames Compare will most often create alarger area of detected motion).
 2 Frames Compare |  3 Frames Compare. |
6) If you decided to interpolate pixels only wherethere is motion, then how do you want to look for motion?
Motion Diff Method:
This setting specifies how a comparison of two frames should bedone in order to determine where there is motion. When PointDifference is selected, corresponding pixels at two differenttimes are differenced. If the difference of the two pixels isgreater than the threshold (specified in the Motion Tolerancesetting, described next), then the motion mask is turned on atthat pixel. Small Area Difference uses a slightly larger region,which will help in reducing problems where video noise shows upas motion in the motion mask (but will also make areas of motionslightly smaller).. Large Area Difference searches an even largerarea at each pixel.
 Motion Diff Method: Point Difference |  Motion Diff Method: Small Area Difference |  Motion Diff Method: Lare Area Difference |
7) Set the Motion Tolerance so that the motionmask is covers the area where there is motion.
Motion Tolerance:0-1: After the difference operation is run, this sliderallows to tweak the region that is turned on in the motion mask.
 lower Motion Tolerance |  higher Motion Tolerance |
After you set up the motion mask region, there are a fewoperators that you can apply to grow and shrink the regions inthe motion mask, along with some noise suppression that can beperformed.
HINT: It helps to set the View Motion Mask to onthen move the Motion Tolerance slider until the motion mask,drawn in the red channel, covers the area of motion.
8) Add some noise suppression to the motiondetection
Noise MaskSuppression Window
This option allows you to delete regions in the motion mask thatare generally the result of video noise. A setting of 0 deletesno regions. A setting of 1 deletes regions that approximately 1pixel in size. A setting of 5 deletes the largest regions. Thejob of this control to remove all the little spurious dots thatare interpreted as motion even though they are just some form ofnoise. Adjust this control until you see no more stand alone dotswhere the image is not moving.
 Mask Suppress set to 0 |  Mask Suppress set to 1 |  Mask Suppress set to 4 |
9) Regrow the valid areas to cover the edgesof objects
Mask Grow:<0-5>
After noise suppression is done, this option grows the motionmask by the specified number of pixels. This can be useful whensetting the threshhold lower includes too many regions, but you'dstill like to grow the motion mask. When you look in close up atthe motion mask image concentrate mostly on the edges of themotion mask and make sure that you do not perceive"combs" outside of the motion mask. If you use 1Frame Detect most probably you will want to make this valueat least the same as the Suppression Window size.
 Mask Grow set to 0 |  Mask Grow set to 3 |
10) Do you what a soft motion mask?
Mask Feather:<0-5>
This parameter create a smooth motion mask by feathering theedges by the number of pixels specified. Feathering can allow youto include a smoother transition between areas where there ismotion detected and areas where there is no motion detected.
 Mask Feather set to 0 |  Mask Feather set to 3 |
11) Now do you want to smooth the areas withinthe motion mask?
Smooth vertically:<slider: 0 to 2.0>
When setting this to a value greater than zero, a verticalsmoothing filter will be applied to the interpolated pixels inthe areas of the motion mask AFTER the fields have beeninterpolated. You should know that our vertical smoothing filteris smart and only filters where there is high contrast (that is,it tries not to make smooth areas more smooth). Note, if you usemotion detection then the vertical smoothing filter will only beapplied inside the motion mask.
 Smooth Vertically set to 0 |  Smooth Vertically set to 1.2 |
12) Finally you havethe option of sharpening up lost details
Sharpen: <slider: 0 to 1.0>
This function is conceptually the opposite of the SmoothVertically function. This option allows you to specify howmuch of a sharpening filter is applied. Sharpening happens afterthe full frame is created. Note that because interpolated pixelsare the result of blending other pixels, these interpolatedpixels will often be smoother than necessary. So thispost-process sharpening filter can be very useful. Hint: as arule of thumb you might consider setting this control to half theSmooth Vertically value you have selected.
 Sharpen set to 0 |  Sharpen set to 0.5 |
Reinterlacer
Interlacing is the process of taking two pictures half a frameapart and interleaving them on a scanline basis, meaning youcreate a single frame from two, talking half the scanlines fromone and half from the other. Reinterlacer is provided as aconvenient tool to interlace frames at differing places in yourworkflow.
Field Order:
Output Type:
 |  |  |  |
Original sequence of 4.
 |  |
Interlaced set of 2 frames (the <HalfFrame Rate or <Keep Frame Rate. options).
 |  |  |  |
=Field Blending option used. Notethe same number of frames in and out. This option blends framesacross fields.
Source Layer :
Note: In After Effects, if there are any filters applied tothe source footage, then "Source Layer" should be apre-composition, otherwise filters applied to the source layerwill not be visible. Other applications may require you to putyour source in containers in order to see other applied filters.
Pulldown
This is a very NTSC process. A pulldown is the traditional waya telecine (a device to transfer 24 FPS film to 30 FPS video)retimes the source material. The formula for 3:2 pulldown turns 4successive input frames into 5 output frames where 2 will beinterlaced..
For example, in WSSWW, the first 4 frames areturned into 5 as shown by below:
 |  |  |  |
4 Input frames at 24 fps
 |  |  |  |  |
5 Output frames at 30fps, converted from 24fpsusing 3:2 pulldown
Note: some applications like Commotion already supportpulldowns as a process you can apply anywhere in your workflowwhile working so that this plugin is not of much use to someonein the Commotion environment.
Field Order:
3:2 Pulldown Mode:
Pulldown Phase:
The Pulldown Phase refer to the action taken on a 5 framesincrement basis from the first frame. Here "W" means"whole frame" and "S" means split frame (aninterlaced frame will be or was made from 2 successive frames)..
Source Layer :
Note: This option determines where the source material islocated. If you are making a sequence longer, then you'll have tomake a longer sequence, apply Pulldown to it, and point to thesource material. (see "
Wait... But Why?
The most difficult thing to understand for someone who is newto this industry is why we still have fields:
So consider this: Fields is a concept inherited fromtelevision/video technology history. The initial reason forfields was a technological difficulty. A clever kludge wasdevised, fields, to address two engineering issues. In anutshell, the scan rate was decided to be made in phase to thepower system frequency (60 MHZ for America and 50 for Europe), asit was evaluated that like this AC line interference effects wereminimized in the reproduced picture. The interlace schemewas selected as a fallback compromise because it was the only wayto overcome the technical limitations of the forties (hello! thisis year 2001). The writers of the initial television standardproposal actually initially proposed 44 full frames a second asminimum to avoid flickering problems and their proposal wasrejected as too expensive for deployment of television in the40s, so then came the Fabulous Fields...
Since then, many equipment manufacturers have invented allkinds of reasons to make us believe that such a thing is actuallya good thing. Bottom line, we are still stocked with it 60 yearsor so later.
Now, have you ever complained about strobing when watching afilm transferred to video? Of course, not. So do we need so manyframes per second? It depends. One functionality forgottenhere is that the film camera process has also an additionalattribute, a shutter. That is a shutter opening causes a motionblur in objects that move in effect reducing the required amountof frames required by our brain to achieve fusion ("criticalflicker fusion"). There is no such concept in videosince by definition video requires to be timed to 1/60th (or1/50) of a second per field. The only way to do shuttering invideo is by dropping resolution (for example dropping a field).
Fields are really a problem when doing visual effects. This isone reason why so much commercials are still shot with film.Fields are also bad to display on a computer monitor and as wellare impossibly hard on compressors as they remove local structurein an image. So the truth is that until we can devisedigi-opto-electronic imaging devices which integrate a full framecapture array (buffer) we might be stocked with 3solutions: shoot with film, do high rate video (60 or overFPS progressive) or do a lot of smart processing in post.
And as an additional informational note, within the standardvideo forms (ATSC), the current solution that creates the lesspost-production problems is 60 progressive FPS video. Thereason is that at that frame rate "motion blur" startsto become a less important factor, because there are nowenough frames per second to have your brain completely do thework. For example if you look at a frame of a Showscan film(60 FPS), you will see very sharp images with almost noblur. This is exactly why our computer display have 70-75HZ (and not 24 for example). Also note that in a film theater,the same frame is actually projected twice (or sometimes 3), thatis there is actually a shuttering process in the film projectionsystem itself, and in effect the projector's shutter is shut halfthe time (for one thing allowing film plane motion to the nextframe).
SO, bottom line, whether we like it or not, we have todeal with it. SO this is about opening another track to let youdeal with fields in a manner where hopefully they come into placetotally at the end of your creative thinking process, and helpyou squeeze out the most quality out of your digital videocamcorders.
Copyright 2001 RE:Vision Effects, Inc.