Comb Filter - PowerPoint by T44oqFA


									Comb Filter

   Y-C separation
   2D comb filter
   3D comb filter
   2D adaptive comb filter
   conclusion
Comb filter (Y-C separator)
   Separate the incoming composite
    video signal into luminance (Y) and
    chrominance (C).

NTSC composite signal

   Video signal prepared for broadcast
    contains two major parts, the
    luminance (Y) and chrominance (C).
Basic concept for Y C separate

   1.Frequency- notch , bandpass filter
    ( in early years ).

   2.Chrominance phase difference-
    comb filter ( today ).
     Luminance & chrominance frequency
Notch/bandpass filter system

   significant overlap , energy go
       Chrominance phase difference
          chroma burst reference and the chroma
           information is reversed by 180 degrees on
           alternate lines (luminance of the system to remain
           constant[ monochrome receivers]).

180 degrees
2D comb filter
   Important concept…picture information (y , c) is
    about the same from line to line .
   Special case( y - c , y + c ).


                                           2c          c
Why we call it a "comb filter"
2D weighting comb filter
   different colors vertically through the picture .
   (1/4,1/2,1/4).
   the transition from one color to another is softened.
Horizontal delay
   analog: glass delay line , CCD .
   digital : memory .
3D comb filter
   compare information in successive frames .(odd)
   perfect separation of Y and C for still frame images
    ( key advantage over the 2D filter ).
   motion or serious color changes , 3D filter may
    perform no better than the 2D filter.
   has to store two frames of video data.
   Vertical detail difference (adjacent
   Motion difference (adjacent frames).

 Adaptive comb filter
1.Detect and select one form.
2.Correction terms.
3.Weighting and sum .
An improved 2D adaptive comb filter

   Overview (mode A,B,C,D)
Mode A (does not perform comb filtering)

   Only when video image has serious
    vertical difference.
   Just a bandpass filter.
Mode B (2D weighting with bandpass filter)

   Normal mode ,Y-C can be separated
    quite well.
Mode C (2D weighting without bandpass filter)

   Use under that 3 lines are so close.
   Can separate better than B.
Mode D (2D simple comb filter)

   Use under that detection(2,3)
Common method for difference detection

   Check Y signal only.
   Switch mode when difference exceeds a
    threshold value.
Detection algorithm
   Normal way (upper).
   Special detection (bottom).
   Use these two detections. (T1,T2,T3)
                Y           Y



Mode A (detection (1,3) > T3)

   Which means the vertical detail
    defference is serious.
   Bandpass filter performs better than
    comb filter.
Mode B (detection (1,3) < T3)&(detection (2,3) > T2)

   The mostly encountered situation.
   Use bandpass filter to eliminate
    little difference between Y.
Mode C (detection (1,3) < T1)&(detection (2,3) < T2)

   Which means 3 lines have almost
    the same Y-C information.
   Mode C better than B.
Mode D                            Y
(T1<detection (1,3) < T3)&(detection (2,3) < T2)

   Which means that detection(2,3)
    dominates.                Mode D

   2D,3D,adaptive 2D.

   Academic : adaptive 2D comb filter
    (less gate count ,more strategy).
   Earn money : 3D/2D switch (best
    performance ,largest area)

To top