|
Edge Detection Through High and Bandpass Filters |
Brought to you by Team Phantom
Cruiser and the Power of Steam
|
|
|
|
| Top: Original image. Center: Image filtered with
5th order Butterworth highpass filter, normalized cutoff frequency .5
Bottom: Image filtered with 27th order Butterworth bandpass filter,
normalized passband .2 - .8. |
 |
 |
Left: Bandpass filter. Right: Highpass
filter. |
If only keeping the low frequencies discards the edges of the
image, then attenuating the low frequencies should only keep the
edges. This is exactly what we find when we apply the above highpass
filter to the image. The filtered image only keeps the edges of
sharpest contrast, such as the curvature around the headlights, the
edges of the windshield, and the space around the doors.
A somewhat more complete view of the car is given by the results of
the bandpass filter. This ignores the very high frequency noise in
the image, and brings out most areas in which a change occurs. This
allows it to find not only edges but faster gradients, such as the
ones hidden in the dark grille on the front of the car, and the
pockets in the hubcaps.
|