Edge detection and Cellular Neural Boolean Filtering
Cellular neural Boolean filtering is nonlinear
spatial-domain filtering reduced to direct processing of
the image binary planes by Boolean functions. The most impressive
representative of this family of filters is the Precise Edge
Detection filter.
In contrast to other edge detection algorithms,
the Precise Edge Detection algorithm, which is based on the
concept of cellular neural Boolean filtering, ensures accurate
detection of all the edges, without being dependent on the
orientation, values of brightness jumps, size of details,
or structure of objects.
The classical edge detection algorithms (such
as Sobel, Prewitt, Laplace, and others) have two common disadvantages:
- They detect the edges corresponding to the large brightness
jumps but usually miss the edges corresponding to the
smaller brightness jumps;
- They do not differentiate between edges that correspond
to upward and downward brightness jumps, and end up detecting
the different edges together, as one combined edge.
As a result, the edges corresponding to
the small brightness jumps often can not be detected, while
those edges that are detected may be bold and inaccurate
because of “double” detection of the edges corresponding
to both upward and downward brightness jumps.
The Precise edge detection algorithm implements
edge detection using three different detectors: the first
one detects edges corresponding to upward brightness jumps;
the second detects edges corresponding to downward brightness
jumps; the third detector is used for the joint detection
of upward and downward brightness jumps. This approach is
objective, since it depends neither on a partial image, nor
on the image statistical characteristics, nor on contrast,
nor on dynamic range. Since the Precise Edge Detection algorithm
is operating via separate processing of the image binary
planes, it is possible to detect edges on the noisy images
by excluding the processing of those binary planes where
noise is primarily concentrated.
Using the same approach, the Precise Edge
Detection algorithm may be used for edge detection by narrow
direction. This kind of detection is far more effective than
corresponding classical methods and it is very important
for detection of the specifically oriented details in the
image.
Another important application of the Precise
Edge Detection algorithm is the Edged Segmentation of the
image. This operation makes it possible to extract the regions
with values of brightness (or levels of colors) that are
equal or close to each other. As a result, the details and
the textures with a complicated configuration may be extracted.
To date the Precise Edge Detection algorithm
has produced the best results for the solution of the edge
detection problems.
Examples
(click on the images to enlarge,
make sure you don't have pop-ups disabled)
Edged Segmentation Example
|
|
Original Image |
Edged Segmentation |
Precise Edge Detection Example
|
|
|
Original Radar Satellite
Image |
Precise Edge Detection
Upward |
Precise Edge Detection Downward |
|