ADS Technologies USBAV-708 INSTANT VIDEOMPX, Руководство пользователя

Страница: 63 / 116

The
Guide
63
So
ur
ce
p
re
-pr
oc
es
sin
g
Sp
a
tia
l f
ilt
er
in
g,
Te
m
po
ra
l f
ilte
ring
Pre-processing consists of two filter types, the first is a spatial filter. The spatial
filter is concerned with each pixel in the frame and the pixels surrounding it.
Noise manifests itself as high frequency changes between the colors of adjacent
pixels. By applying a low-pass filter (one that rejects high frequencies) to an
image it is possible to stabilize the colors.
The second type of filter employed by the pre-processing algorithm is a temporal
filter. While the spatial filter can reduce noise within a single frame it does not
consider the changes in each pixel over time, I.e. from frame-to-frame. A good
example of temporal artifacting is flickering video, where even though there may
be little noise within each individual frame there is
temporal noise affecting the
intensity of pixels over time. Temporal noise is also responsible for producing
grain or discoloration patterns that change throughout a video as it is played.
Just as it is possible to perform low-pass filtering within a frame by examining
adjacent pixels it is also possible to consider the color of each individual pixel
with respect to its color in several previous frames. By applying an adaptive low-
pass filter to co-located pixels throughout time any temporal noise related to
lighting or color tonality between frames can be reduced.
Spatial filtering
Noisy image Sampled data After low-pass filter Filtered image
Low-pass spatial filtering
Temporal filtering
Temporal filtering