View source for TDeint

{{FilterCat4|External_filters|Plugins|Restoration_filters|Deinterlacers}}
{{Filter3
| {{Author/tritical}}
| v1.1
| [http://web.archive.org/web/20140420182314/http://bengal.missouri.edu/~kes25c/TDeintv11.zip TDeintv11.zip]
| Deinterlacing 
| [http://www.gnu.org/licenses/gpl-2.0.txt GPLv2]
|6=[http://forum.doom9.org/showthread.php?t=82264 Doom9 Thread]
}}

== Description ==
TDeint is a bi-directionally, motion adaptive, sharp deinterlacer. It can adaptively choose between using per-field and per-pixel motion adaptivity, and can use cubic interpolation, kernel interpolation (with temporal direction switching), or one of two forms of modified ELA interpolation which help to reduce "jaggy" edges in moving areas where interpolation must be used. TDeint also supports user overrides through an input file, and can act as a smart bobber or same frame rate deinterlacer, as well as an IVTC post-processor.
<br>
<br>
== Requirements ==
* AviSynth 2.5.8 or later
* Supported color formats: [[YUY2]], [[YV12]]
<br>

== [[Script variables|Syntax and Parameters]] ==
:{{Template:FuncDef|TDeint (clip, int "mode", int "order", int "field", int "mthreshL", int "mthreshC", int "map", string "ovr", int "ovrDefault", int "type", bool "debug", int "mtnmode", bool "sharp", bool "hints", PClip "clip2", bool "full", int "cthresh", bool "chroma", int "MI", bool "tryWeave", int "link", bool "denoise", int "AP", int "blockx", int "blocky", int "APType", PClip "edeint", PClip "emask", float "blim", int "metric", int "expand", int "slow", PClip "emtn", bool "tshints", int "opt")}}
<br>


{{Par2| |clip| }}
:Input clip.
<br>
<p><var>mode</var>:</p>
<ul>
<p>Sets the mode of operation.  Modes -2 and -1 require progressive input.</p>
<ul>
<li>-2 - double height using modified ELA
<li>-1 - double height using modified ELA-2
<li>&nbsp;0 - same rate output
<li>&nbsp;1 - double rate output (bobbing)
<li>&nbsp;2 - smartbobbed field-matching (same rate output, blend frames from bobbed stream)
</ul>
<p>default -&nbsp;&nbsp;0  (int)</p>
</ul>


<p><var>order</var>:</p>
<ul>
<p>Sets the field order of the video.</p>
<ul>
<li>-1 - use parity from Avisynth
<li>&nbsp;0 - bottom field first (bff)
<li>&nbsp;1 - top field first (tff)
</ul>
<p>default -&nbsp;&nbsp;-1  (int)</p>
</ul>


<p><var>field</var>:</p>
<ul>
<p>When in mode 0 and 2, this sets the field to be interpolated.  When in mode 1, this 
setting does nothing.</p>
<ul>
<li>-1 - will set field equal to order if hints = false or to 0 if hints = true
<li>&nbsp;0 - interpolate top field (keep bottom field)
<li>&nbsp;1 - interpolate bottom field (keep top field)
</ul>
<p>default -&nbsp;&nbsp;-1  (int)</p>
</ul>


<p><var>mthreshL</var>/<var>mthreshC</var>:</p>
<ul>
<p>The motion thresholds for luma and chroma (mthreshL for luma, mthreshC for chroma). If 
the difference between two pixels is less than this value they are declared static. 
Smaller values will reduce residual combing, larger values will decrease flicker and 
increase the accuracy of field construction in static areas.  The spatially corresponding 
parts of the luma and chroma planes are linked (if link != 0), so mthreshC and mthreshL 
may be somewhat interconnected.  Setting both values to 0 or below will disable motion 
adaptation (i.e. every pixel will be declared moving) allowing for a dumb bob.</p>
<p>default -&nbsp;&nbsp;<var>mthreshL</var> - 6  (int)<br>
&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;<var>mthreshC</var> - 6  (int)</p>
</ul>


<p><var>map</var>:</p>
<ul>
<p>Displays an output map instead of the deinterlaced frame.  There are five possible options:</p>
<ul>
<li>0 - No map.
<li>1 - value (binary) map.  This will output a frame in which all the pixels have one of the
        following values (indicating how the frame is to be constructed):
                 <ul>
                 <li>0   (use pixel from current frame)
                 <li>51  (use pixel from previous frame)
                 <li>102 (use pixel from next frame)
                 <li>153 (use average of curr/next)
                 <li>204 (use average of curr/prev)
                 <li>230 (use [1 2 1] average of prev/curr/next)
                 <li>255 (interpolate pixel)
                 </ul>
<li>2 - Merged map.  This will output a frame in which all the static parts of the frame (values 
        0, 51, 102, 153, 204, and 230 from map=1) have been constructed as they would appear in the 
        deinterlaced frame, and the pixels that are to be interpolated are marked in white (255).
<li>3 - Outputs the deinterlaced frame and the value (binary) map stacked vertically (the map is on the bottom).
<li>4 - Outputs the deinterlaced frame and the merged map stacked vertically (the map is on the bottom).
</ul>
<p>default -&nbsp;&nbsp;0  (int)</p>
</ul>


<p><var>ovr</var>:</p>
<ul>
<p>Sets the name and path to an overrides file.  When mode=0, an overrides file can be used to 
control the values of <var>mthreshL</var>, <var>mthreshC</var>, <var>field</var>, <var>order</var>, and 
<var>type</var> for single frames or for ranges of frames, as well as control which frames are 
deinterlaced. When mode=1, an overrides file can be used to control the values of <var>mthreshL</var>,
<var>mthreshC</var>, and <var>type</var> for specific frames or ranges of frames.</p>
<p>Overrides file specifiers:</p>
<ul>
<li>+ = mark frame to be deinterlaced (only useful if ovrDefault = 1)
<li>- = mark frame to not be deinterlaced
<li>f = field
<li>o = order
<li>l = mthreshL
<li>c = mthreshC
<li>t = type
<p>*The c, f, o, l, t specifiers also require a change value to be specified when they are used 
(look at the overrides syntax to see how this is done)</p>
</ul>
<p>Override syntax:</p>
<ul>
<p>&nbsp;&nbsp;&nbsp;&nbsp;[] = not required for +, - specifiers</p><br>
<DT>single frame override:
<DD><br>frame_number specifier [change_value]
  <DD><br>examples:<ul>
<pre>
245 f 1
345 +
400 -
450 c -1
</pre></ul><br>
<DT>override for range of frames:
<DD><br>start_frame_number,end_frame_number specifier [change_value]
  <DD><br>examples:<ul>
<pre>
100,200 +
346,352 f 0
900,1200 l 5
</pre></ul><br>
<DD>** The range is inclusive, meaning the end frame and start frame are both included.
<DD><br>Pattern based frame range overrides (only for +,- specifiers):
<DD><br>examples:<ul>
<pre>
100,300 +-+++--+++
400,456 ---+---++
</pre></ul><br>
<DD>** Will use the given pattern over the specified frame range.
</ul>
<p>Things to remember (key points/rules):</p>
<ul>
<ol>
  <li>Ranges are inclusive
  <li>When mode = 1 (bobbing) all overrides except for mthreshL/mthreshC, and type overrides are ignored. 
        Also, frame #'s correspond to the input clip not the output clip, thus one frame will be two frames 
        in the output.
  <li>The changed value is always set back to what it was originally set to after the override goes out 
        of the specified range. (i.e. if you specify an mthresh override for frame 600 to 700 after 
        frame 700 mthresh is set back to its original value automatically, you don't need to set it back 
        in the overrides file!  The original value is what it is set to on load (i.e. either the default 
        or what you set it to in your avisynth script).
  <li>Frame numbers must be within range for the file.
  <li>Frame numbers for specific specifiers must be ascending (if they are not, the last entry in the 
        file takes precedence ex. if you specify 300,400 c 10 then later do 350,450 c 12 frames 350 to 400 
        will use 12 not 10).
  <li>Frames numbers for the (+, -) specifiers cannot overlap (e.g. don't do 300,400 - and then later in the 
        file write 350,500 + or strange things will happen.  The other specifiers don't have to meet this 
        requirement as they all effect different things.
  <li>+, - specifiers require no change value.
  <li>The spacing is important! Just look at the examples.
  <li>Only +, - specifiers can be used in pattern specifications.
  <li>You can change multiple specifiers over the same frame range as long as you follow the rules above 
        (+, - ascending frame numbers for example).
  <li>You can comment out a line (i.e. it will be ignored) by adding a '#' or ';' to the beginning of the line.
  <li>Entering 0 as the end_frame for a range of frames is taken as meaning the last frame of the video.
</ol>
</ul>
<p>Example overrides file:</p>
<ul>
<p>Syntax example => TDeint(order=1,ovr=&quot;c:\path\myoverridesfile.txt&quot;)</p>
<pre>
100,300 o 0
100,300 f 1
90,250 c 3
40,500 -
505 -
300,700 l -1
#700,3000 f 1 &lt;- commented out, will be ignored
800,1000 -++-
500,1000 c 13
</pre>
</ul>
<p>default -&nbsp;&nbsp;""  (string)</p>
</ul>


<p><var>ovrDefault</var>:</p>
<ul>
<p>When using an overrides file in mode 0, this specifies the default action for all frames in the video. 
Using ovrDefault=1 makes it easy to deinterlace only a few specific frames in a video.  When mode = 1, 
this setting does nothing.</p>
<ul>
<li>0 - all frames not specified as '-' in the overrides file are deinterlaced
<li>1 - all frames not specified as '+' in the overrides file are not deinterlaced and simply returned as is
</ul>
<p>default -&nbsp;&nbsp;0  (int)</p>
</ul>


<p><var>type</var>:</p>
<ul>
<p>Sets the type of interpolation to use.  Cubic is the fastest, modified ELA and ELA2 will give 
smoother, less "jaggy", edges and are the slowest (ELA2 is faster), and kernel interpolation will 
cause significantly less flickering than cubic or ela when interpolation gets used in almost static 
areas.  Modified ELA and ELA2 work best with anime/cartoon type material... they are not that great with 
real life sources (sometimes they are, test for yourself).</p>
<ul>
<li>0 - cubic interpolation
<li>1 - modified ELA interpolation
<li>2 - kernel interpolation (can be normal or sharp, controlled by the sharp setting)
<li>3 - modified ELA-2 interpolation
<li>4 - blend interpolation
<li>5 - bi-directional blend interpolation
</ul>
<p>default -&nbsp;&nbsp;2  (int)</p>
</ul>


<p><var>debug</var>:</p>
<ul>
<p>Will enable debug output, which for each frame will list the values of <var>order</var>, <var>field</var>, 
<var>mthreshL</var>, <var>mthreshC</var>, and <var>type</var> if the frame is being deinterlaced.  If the 
frame is not being deinterlaced (due to user overrides, <var>hints</var>, or <var>full</var>=false), it will 
simply say the frame is not being deinterlaced and list the specific reason. If the output frame is weaved, 
then debug output will report which field the current field was weaved with (PREV or NEXT).  The debug 
information is output using OutputDebugString().  To view the output you can use [http://technet.microsoft.com/en-us/sysinternals/bb896647 DebugView] from Sysinternals.</p>
<p>default -&nbsp;&nbsp;false  (bool)</p>
</ul>


<p><var>mtnmode</var>:</p>
<ul>
<p>Controls whether a 4 field motion check or a 5 field motion check is used.  5 field will prevent more 
artifacts and can deal with duplicate interlaced frames; however, it is quite a bit slower than the 4 field 
motion check.  Modes 2 and 3 are like 0 and 1 except that in areas where an average of the prev and next 
field would have been used in mode 0 or 1, the pixel value from the most similar field (computed via field 
differencing) is used instead (i.e. no averages are used).</p>
<ul>
<li>0 - 4 field check
<li>1 - 5 field check
<li>2 - 4 field check (no averages, replace with most similar field)
<li>3 - 5 field check (no averages, replace with most similar field)
</ul>
<p>default -&nbsp;&nbsp;1  (int)</p>
</ul>


<p><var>sharp</var>:</p>
<ul>
<p>Controls whether the sharp or normal kernel is used when using kernel interpolation (<var>type</var> = 2). 
The sharp kernel includes more pixels and produces a sharper result but is slightly slower.</p>
<ul>
<li>true - use sharp kernel
<li>false - use normal kernel
</ul>
<p>default -&nbsp;&nbsp;true  (bool)</p>
</ul>


<p><var>hints</var>:</p>
<ul>
<p>Read hints from telecide or tfm indicating which frames are interlaced and which are not if hints 
are present in the video stream.  To make this work you need to set post=1 in telecide or PP=1 in tfm 
and put TDeint immediately afterwards.  TDeint will not effect the hints (as long as your video has a 
width of at least 64 pixels) in case any filters later on need to read them.  If hints is set to true, 
but no hints from telecide or tfm are detected in the video stream, then all frames will be deinterlaced 
(TDeint will operate as if hints=false).  If you do not specify a value for hints explicitly, then 
TDeint will check to see if hints are present in the stream on load and set hints to true if they are 
or false if they aren't (i.e. it is automatically set).</p>
<p>**NOTE:  for IVTC post-processing by reading hints it is recommended to use TDeint in
the following fashion making use of the clip2 parameter.</p>
<ul>
<pre>
orig = last
fieldmatcher()
TDeint(clip2 = orig)
</pre><br>
<li>true - read hints if present
<li>false - don't read hints
</ul>
<p>default -&nbsp;&nbsp;automatically detected on load  (bool)</p>
</ul>


<p><var>clip2</var>:</p>
<ul>
<p>If using tdeint as a postprocessor for telecide or tfm via the hints parameter (or any field matcher), 
incorrect deinterlacing can occur due to the fact that telecide changes the order of the fields in the 
original stream (it is a field matcher after all).  This can cause problems in some cases since TDeint 
really needs to have the original stream.  To work around this, you can specify a second clip "clip2" 
for TDeint to do the actual deinterlacing from.</p>
<p>In a script this is how it would work:</p>
<pre>
mpeg2source(&quot;c:\mysource.d2v&quot;)
orig = last
telecide(guide=1, order=1, hints=true, post=1)
tdeint(order=1, clip2=orig)
</pre>
<p>So TDeint reads the output clip from telecide as usual.  When hints indicate an interlaced
frame, it does the deinterlacing of the frame using clip2.  This method also perserves the hints 
in the output stream so any other filters that need them later on will still work.</p>
<p>With the addition of full=false, another way to use TDeint as a post-processor is to
have it use its own combed frame detection as follows (this also allows it to work with any field matcher, 
not just telecide or tfm):</p>
<pre>
mpeg2source(&quot;c:\mysource.d2v&quot;)
orig = last
fieldmatcherofchoice()
tdeint(order=1, full=false, clip2=orig)
</pre>
<p>default -&nbsp;&nbsp;NULL  (PClip)</p>
</ul>


<p><var>full</var>:</p>
<ul>
<p>If full is set to true, then all frames are processed as usual.  If full=false, all frames
are first checked to see if they are combed.  If a frame isn't combed, then it is returned as is. 
If a frame is combed, then it is processed as usual.  The parameters that effect combed frame 
detection are <var>cthresh</var>, <var>chroma</var>, <var>blockx</var>, <var>blocky</var>, and 
<var>MI</var>. full=false allows TDeint to be an ivtc post-processor without the need for hints.</p>
<ul>
<li>true - normal processing
<li>false - check all input frames for combing first
</ul>
<p>default -&nbsp;&nbsp;true  (bool)</p>
</ul>


<p><var>cthresh</var>:</p>
<ul>
<p>Area combing threshold used for combed frame detection.  It is like dthresh or dthreshold
in telecide() and fielddeinterlace().  This essentially controls how "strong" or "visible" combing 
must be to be detected.  Good values are from 6 to 12. If you know your source has a lot of combed 
frames set this towards the low end (6-7). If you know your source has very few combed frames set 
this higher (10-12).  Going much lower than 5 to 6 or much higher than 12 is not recommended.</p>
<p>default -&nbsp;&nbsp;6  (int)</p>
</ul>


<p><var>blockx</var>:</p>
<ul>
<p>Sets the x-axis size of the window used during combed frame detection.  This has to do with 
the size of the area in which <var>MI</var> number of pixels are required to be detected as combed for 
a frame to be declared combed.  See the <var>MI</var> parameter description for more info.  Possible 
values are any number that is a power of 2 starting at 4 and going to 2048 (e.g. 4, 8, 16, 32, ... 2048).</p>
<p>default -&nbsp;&nbsp;16  (int)</p>
</ul>


<p><var>blocky</var>:</p>
<ul>
<p>Sets the y-axis size of the window used during combed frame detection.  This has to do with 
the size of the area in which <var>MI</var> number of pixels are required to be detected as combed for 
a frame to be declared combed.  See the <var>MI</var> parameter description for more info.  Possible 
values are any number that is a power of 2 starting at 4 and going to 2048 (e.g. 4, 8, 16, 32, ... 2048).</p>
<p>default -&nbsp;&nbsp;16  (int)</p>
</ul>


<p><var>chroma</var>:</p>
<ul>
<p>Includes chroma combing in the decision about whether a frame is combed.  Only use this if 
you have one of those weird sources where the chroma can be temporally separated from the luma 
(i.e. the chroma moves but the luma doesn't in a field).  Otherwise, it will just help to screw 
up the decision most of the time.</p>
<ul>
<li>true - include chroma combing
<li>false - don't
</ul>
<p>default -&nbsp;&nbsp;false  (bool)</p>
</ul>


<p><var>MI</var>:</p>
<ul>
<p>The number of required combed pixels inside any of the <var>blockx</var> by <var>blocky</var> sized 
blocks on the frame for the frame to be considered combed.  While <var>cthresh</var> controls how "visible" 
or "strong" the combing must be, this setting controls how much combing there must be in any localized area (a 
<var>blockx</var> by <var>blocky</var> sized window) on the frame.  Min setting = 0, max setting = 
<var>blockx</var> x <var>blocky</var> (at which point no frames will ever be detected as combed).</p>
<p>default -&nbsp;&nbsp;64  (int)</p>
</ul>


<p><var>tryWeave</var>:</p>
<ul>
<p>If set to true, when TDeint deinterlaces a frame it will first calculate which field (PREV or NEXT) 
is most similar to the current field.  It will then weave this field to create a new frame and check 
this new frame for combing.  If the new frame is not combed, then it returns it. If it is, then it 
deinterlaces using the usual per-pixel motion adaptation.  Basically, this setting allows TDeint to 
try to use per-field motion adaptation instead of per-pixel motion adaptation where possible.</p>
<p>default -&nbsp;&nbsp;false  (bool)</p>
</ul>


<p><var>link</var>:</p>
<ul>
<p>Controls how the three planes (Y, U, and V) are linked during comb map creation. Possible settings:</p>
<ul>
<li>0 - no linking
<li>1 - Full linking (each plane to every other)
<li>2 - Y to UV (luma to chroma)
<li>3 - UV to Y (chroma to luma)
</ul>
<p>default -&nbsp;&nbsp;2  (int)</p>
</ul>


<p><var>denoise</var>:</p>
<ul>
<p>Controls whether the comb map is denoised or not.  True enables denoising, false disables.</p>
<p>default -&nbsp;&nbsp;false  (bool)</p>
</ul>


<p><var>AP</var>:</p>
<ul>
<p>Artifact protection threshold.  If AP is set to a value greater than or equal to 0, then before 
outputting a deinterlaced frame TDeint will scan all weaved pixels to see if any create a value 
greater than AP.  Any pixels that do will be interpolated.  Use this to help prevent very obvious 
motion adaptive related artifacts. A large value for AP is recommended (25+, or as large as removes 
the artifacts that can be seen during full-speed playback), as smaller values will destroy the 
benefits of motion adaptivity in static, detailed areas.  The AP metric is the same as the <var>cthresh</var> 
metric... so the scale is 0-255. At zero everything but completely flat areas will be detected as 
combing. At 255 nothing will be detected. Using AP will slow down processing. Set AP to a value less 
than 0 or greater than 254 to disable.</p>
<p>default -&nbsp;&nbsp;-1 (disabled)  (int)</p>
</ul>


<p><var>APType</var>:</p>
<ul>
<p>When <var>AP</var> post-processing is being used (<var>AP</var> is set >= 0 and < 255), APType 
controls whether the motion of surrounding pixels should be taken into account.  There are 3 
possible settings:</p>
<ul>
<li>0 = Don't take surrounding motion into account.  If a weaved pixel creates a value that
        exceeds the AP threshold then it will be interpolated.
<li>1 = If a weaved pixel creates a value that exceeds the AP threshold and at least half of
        pixels in a 5x5 window centered on that pixel were detected as moving then that
        pixel will be interpolated.
<li>2 = Exactly like 1, except instead of 1/2 only 1/3 of the pixels in the surrounding 5x5
        window must have been detected as moving.
</ul>
<p>Modes 1 and 2 provide a way to catch more artifacts (low AP values) without completely
sacrificing static areas.</p>
<p>default -&nbsp;&nbsp;1  (int)</p>
</ul>


<p><var>edeint</var>:</p>
<ul>
<p>Allows the specification of an external clip from which to take interpolated pixels instead 
of having TDeint use one of its internal interpolation methods.  If a clip is specified, then 
TDeint will process everything as usual except that instead of computing interpolated pixels 
itself it will take the needed pixels from the corresponding spatial positions in the same frame 
of the edeint clip.  To disable the use of an edeint clip simply don't specify a value for edeint.</p>
<p>default -&nbsp;&nbsp;NULL (PClip)</p>
</ul>


<p><var>emask</var>:</p>
<ul>
<p>Allows the specification of an external clip from which to take the motion mask instead of 
having TDeint build the mask internally.  Using this option makes the following parameters of 
TDeint have no effect:  <var>mthreshL</var>, <var>mthreshC</var>, <var>mtnmode</var>, <var>denoise</var>, 
<var>link</var>.  The possible values that can be present in the motion mask frames are defined as 
follows:</p>
<ul>
<li>10 - Use pixel from current frame
<li>20 - Use pixel from previous frame
<li>30 - Use pixel from next frame
<li>40 - Use avg of pixels from current and next
<li>50 - Use avg of pixels from current and previous
<li>60 - Interpolate
<li>70 - Use [1 2 1] blend of pixels from prev/curr/next
</ul>
<p>Behavoir is undefined for other values, but they should end up being treated internally 
as though they were 60.</p>
<p>default -&nbsp;&nbsp;NULL (PClip)</p>
</ul>


<p><var>blim</var>:</p>
<ul>
<p>Sets the maximum difference value for mode 2.  If both differences (src-prev and src-next) are 
above this value then src is returned as is.  Otherwise, src is blended with either prev or next 
depending on which is most similar to src.  This value is on a 0.0 to 100.0 scale based on luma plane 
difference.  Use debug=true to see the difference values generated and the limit value.  The debug 
output will look like the following:</p>
<pre>[5776] TDeint:  frame 0:  d1 = 0  d2 = 0  lim = 1513728</pre>
<p>d1 is the src-prev difference and d2 is the src-next difference.  lim is the maximum value 
translated from the float value into an unsigned long value.  Set blim to a negative value to 
disable checking (src will always be blended with either prev or next).</p>
<p>default -&nbsp;&nbsp;-2.0  (float)</p>
</ul>


<p><var>metric</var>:</p>
<ul>
<p>Sets which spatial combing metric is used to detect combed pixels.  Possible options:</p>
<ul>
<p>Assume 5 neighboring pixels (a,b,c,d,e) positioned vertically.</p>
<ul><p>
a<br>
b<br>
c<br>
d<br>
e<br>
</p></ul>
<pre>
0:  d1 = c - b;
    d2 = c - d;
    if ((d1 > cthresh && d2 > cthresh) || (d1 < -cthresh && d2 < -cthresh))
    {
       if (abs(a+4*c+e-3*(b+d)) > cthresh*6) it's combed;
    }

1:  val = (b - c) * (d - c);
    if (val > cthresh*cthresh) it's combed;
</pre>
</ul>
<p>Metric 0 is what tdeint always used previous to v1.0 RC7.  Metric 1 is the combing metric 
used in Donald Graft's FieldDeinterlace()/IsCombed() funtions in decomb.dll.</p>
<p>default -&nbsp;&nbsp;0  (int)</p>
</ul>


<p><var>expand</var>:</p>
<ul>
<p>Sets the number of pixels to expand the comb mask horizontally on each side of combed pixels. 
Basically, if  expand is greater than 0 then TDeint will consider all pixels within 'expand' distance 
horizontally of a detected combed pixel to be combed as well.</p>
<p>default -&nbsp;&nbsp;0  (int)</p>
</ul>


<p><var>slow</var>:</p>
<ul>
<p>Sets which field matching function is used.  These functions match the corresponding functions 
in tfm.  Possible values:</p>
<ul>
<li>0 - normal  (should have the worst accuracy)
<li>1 - slower
<li>2 - slowest (should have the best accuracy)
</ul>
<p>default -  1&nbsp;&nbsp;(int)</p>
</ul>


<p><var>emtn</var>:</p>
<ul>
<p>Specifies an external clip that TDeint will use for motion detection instead of running it on the 
input clip.  The input clip is still used for deinterlacing.  The <var>emtn</var> clip is simply used for 
calculating pixel changes between frames.</p>
<p>default -  NULL&nbsp;&nbsp;(PClip)</p>
</ul>


<p><var>tshints</var>:</p>
<ul>
<p>Enables outputting of hints for use by TSwitch().  If TSwitch is being used after TDeint then set this to 
true.  Otherwise, set this to false.</p>
<p>default -  false&nbsp;&nbsp;(bool)</p>
</ul>


<p><var>opt</var>:</p>
<ul>
<p>Controls which cpu optimizations are used.  Possible settings:</p>
<ul>
<li>0 - use c routines
<li>1 - use mmx routines
<li>2 - use isse routines
<li>3 - use sse2 routines
<li>4 - auto detect
</ul>
<p>default -&nbsp;&nbsp;4  (int)</p>
</ul>

<br>
===TSwitch===
TSwitch is able to read hints from TDeint indicating which frames were deinterlaced or not deinterlaced. Based on these hints, TSwitch then delivers frames from one of two clips (c1 or c2). If the input frame was not deinterlaced, then the corresponding frame from clip c1 is returned. If the input frame was deinterlaced, then the corresponding frame from clip c2 is returned.

To use TSwitch, you must set tshints=true in TDeint and place TSwitch immediately after TDeint in your Avisynth script.

The reasons why a frame may not have been deinterlaced by TDeint are overrides, hints, full=false, or tryweave=true.
<br>
<br>
:{{Template:FuncDef|TSwitch (clip, clip "c1", clip "c2", bool "debug")}}
<br>
<p><var>c1</var>:</p>
<ul>
<p>Sets the clip to return frames from when the input frame was not deinterlaced.</p>
<p>default -  NULL&nbsp;&nbsp;(PClip)</p>
</ul>

<p><var>c2</var>:</p>
<ul>
<p>Sets the clip to return frames from when the input frame was deinterlaced.</p>
<p>default -  NULL&nbsp;&nbsp;(PClip)</p>
</ul>

<p><var>debug</var>:</p>
<ul>
<p>Enables debug output, which for each frame will say whether or not the input frame was deinterlaced and 
which clip (<var>c1</var> or <var>c2</var>) the output frame is being taken from.  The debug 
information is output using OutputDebugString().  To view the output you can use 
[http://technet.microsoft.com/en-us/sysinternals/bb896647 DebugView] from Sysinternals.</p>
<p>default -&nbsp;&nbsp;false  (bool)</p>
</ul>
<br>
<br>

== Examples == 
<p><b>Same rate deinterlacing:</b></p>
<ul><pre>mpeg2source()
tdeint()</pre></ul>

<p><b>Bobbing:</b></p>
<ul><pre>mpeg2source()
tdeint(mode=1)</pre></ul>

<p><b>Deinterlacing with EEDI2 for interpolation:</b></p>
<ul><pre>mpeg2source()
interp = separatefields().selecteven().eedi2()
tdeint(edeint=interp)</pre></ul>

<p><b>Bobbing with EEDI2 for interpolation:</b></p>
<ul><pre>mpeg2source()
interp = separatefields().eedi2(field=-2)
tdeint(mode=1,edeint=interp)</pre></ul>

<p><b>Smartbobbed field-matching (same rate deinterlacing via blending of bobbed frames):</b></p>
<ul><pre>mpeg2source()
tdeint(mode=2)</pre></ul>

<p><b>Smartbobbed field-matching with EEDI2 for interpolation:</b></p>
<ul><pre>mpeg2source()
interp = separatefields().eedi2(field=-2)
tdeint(mode=2,edeint=interp)</pre></ul>

<p><b>Using TSwitch to sharpen only deinterlaced frames:</b></p>
<ul><pre>mpeg2source()
tdeint(tshints=true)
tswitch(c1=last,c2=last.sharpen())</pre></ul>

<br>
== Changelog ==
 Version      Date            Changes<br>
<br>
== Archived Downloads ==
{| class="wikitable" border="1"; width="400px"
|-
!!width="100px"| Version
!!width="150px"| Download
!!width="150px"| Mirror
|-
!v1.7
|<strike>[http://bengal.missouri.edu/~kes25c/TDeintv11.zip TDeintv11.zip]</strike>
|[http://web.archive.org/web/20141216072217/http://bengal.missouri.edu/~kes25c/TDeintv11.zip TDeintv11.zip]
|}
<br>
== External Links ==
*[http://forum.doom9.org/showthread.php?t=82264 Doom9 Forum] - TDeint discussion.
<br>
<br>
-----------------------------------------------
'''Back to [[External_filters#Deinterlacing|External Filters]] &larr;'''