Professional Documents
Culture Documents
KEYWORDS:
Edgel: The term edgel is used to represent a single high contrast gradient point to avoid
confusion with the topological object called an edge.
gradient: vector whose first component is the direction in which the first derivative is highest
and whose second component is the magnitude of the first derivative in that direction.
Convolution : Point wise multiplication of each kxk neighborhood of a digital image by a kxk
kernel of convolution weights
Distance Transform: Given a binary image, its distance transform is an image in which each
pixel is assigned its distance from the nearest 1.
chamfer matching: Convolution of a binary edge template with a chamfer image for matching.
Junction: A junction is a 1 pixel in a thinned binary image that has more than 2 neighbors that
are 1's.
• Initial processing
• Sub-pixel localization
• Thresholding of N
• Edgel Thinning
• Retrieving the topological description
Initial processing
This step is same as canny edge detector.
The intensity image I is smoothed with a discretely sampled two dimensional isotropic Gaussian
Sub-pixel localization
For each pixel having N>t1 where t1 is a pre-set threshold we can estimate where an actual
edgel would lie using the sub-pixel interpolation technique suggested by Canny
Thresholding of N:
In rothwell edge detector only one threshold is used .Itdo not make use of a threshold because
there is no reason to believe that the strength of an event has any particular relevance to its value
for a higher level process such as object recognition.It is quite likely that illumination effects and
shadows can reduce the contrast of an object boundary .
Edgel thinning
Thresholding yields an edgel image which has connected subsets. Two elements of are
immediately connected if they are 8-way neighbours of each other.It uses chamfer convolution
and distance transform matrix.
CHAMFER CONVOLUTION:
Given:
binary image, B, of edge and local feature location binary “edge template”, T, of shape we
want to match
Let D be an array in registration with B such that D(i,j) is the distance to the nearest “1” in
B. this array is called the distance transform of B
Goal: Find placement of T in D that minimizes the sum, M, of the distance transform multiplied
by the pixel values in T if T is an exact match to B at location (i,j) then M(i,j) = 0 but if the
edges in B are slightly displaced from their ideal locations in T, we still get a good match using
the distance transform technique
Chamfer matching is convolution of a binary edge template with the distance transform ,for any
placement of the template over the image.It sums up the distance transform values for all pixels
that are “1’s” (edges) in the template.
If, at some position in the image, all of the edges in the template coincide with edges in the
image (which are the points at which the distance transform is zero), then we have a perfect
match with a match score of 0.
(ROTHWELL)
ADVANTAGES OF ROTHWELL:
• Subpixel accuracy of the detected edges
• Better performance at edge junctions
REFERENCES:
• Dr.Oh Class Web site http://www.cse.unt.edu/~jhoh.
• Driving Vision by Topology
• Evaluating edge detection through boundary detection
• More on template matching Chamfer matching Computing the distance ...