29-05-2012, 12:51 PM
An Improved Canny Algorithm for Edge Detection
An Improved Canny Algorithm for Edge Detection.pdf (Size: 273.13 KB / Downloads: 1,200)
Introduction
Edge detection is an important part of the digital image processing. The edge is the set of the pixel, whose
surrounding gray is rapidly changing. The internal characteristics of the edge-dividing area are the same,
while different areas have different characteristics. The edge is the basic characteristics of the image. There
is a lot of information of the image in the edge. Edge detection is to extract the characteristics of discrete
parts by the difference in the image characteristics of the object, and then to determine the image area
according to the closed edge. Edge detection is widely used in computer vision, image analysis, etc.
Edge detection methods are mainly as follows:
1) Edge detection based on gradient operator.
The edge is the place where image gray value is changing rapidly, so the method based on the derivation
of the gradient operator is most widely used. The classical gradient operators are Sobel operator[1], Prewitt
operator[2], Roberts operator, Laplacian operator.
Canny Optimum Criteria
Canny believed that a well-performing edge detection operator should have the following three
characteristics:
a) Low probability on mismarking non-edge points and low probability on nonmarking the real edge
points.
b) The pixel marked as the edge point should be as close to the center of the real edge as possible.
c) The result of applying the operator is that there is only one pixel marked as the edge point.
According to these characteristics, Canny put forward that the edge detection operator should satisfy
three optimal criteria: signal to noise ratio (SNR) criterion, location accuracy criterion and single-edge
response criterion. And then he suggested the optimal edge detection algorithm.
Canny Edge Detection
Canny operator is the optimum-approaching operator of the product of SNR and the location. Canny
algorithm smoothes image by Gaussian filter, calculates the magnitude and direction of gray level gradient,
has the non-maxima suppression on gradient magnitude, and detect and connect the edge from the
candidate points by the high and low thresholds. Figure 1 shows the basic steps of Canny algorithm.
The Non-maxima Suppression on the Gradient Magnitude
In order to determine the edge of the image, the roof ridge of gradient magnitude image shall be refined.
Only the local maximum of the magnitude shall be kept, that is, non-maxima shall be suppressed to get the
refined edge. Canny operator has the interpolation, along the gradient direction, in the gradient magnitude
imageG in the 2×2 neighborhood of the center point(i, j) . If the gradient magnitude of the pointM(i, j) is
greater than the two adjacent interpolation in the direction ofθ (i, j) , the point(i, j)will be marked as the
candidate edge point, Otherwise marked as non-edge point.
Improved Method of Gradient Magnitude and Direction
The traditional Canny operator calculates, in the neighborhood of 2× 2 , the difference the gradient
magnitude, the gradient directions are horizontal, vertical, left diagonal and right diagonal zones. This
method is more sensitive to noise. The non-edge could be detected and the real edge could be missed. We
presents a gradient magnitude in the 8 neighborhood, which can effectively suppress noise and precisely
locate the edge.