Edge detector performance (Pratt's Figure of Merit)

Hi,

I am trying to compare my edge detection algorithm with other standard ones (Sobel,Canny etc). I found one metric which can be used to quantify the performance of edge detector i.e. Pratt's FOM http://homepages.inf.ed.ac.uk/rbf/CVonline/LOCAL_COPIES/OWENS/LECT6/node2.html

I am unable to understand few things 1) What does "distance between the actual and the ideal edges" mean and how to implement this logic in MATLAB ?

(Somewhere from internet I came to know that actual edge is the edge which I am getting from my algo and ideal edge is the edge from some standard edge detection algo which can be considered as ideal. Consider a=1/9).

2)Why this metric is running in single summation means how it is possible to calculate this metric for whole image (which requires two summation) by using single summation only.

Please explain in terms of pseudocode. Thanking in advance.

Vivek

Answer 1

[David Young]

The Pratt measure requires a "gold standard" - that is, something that gives you perfect or ideal edges for your images. You could use a standard edge detection algorithm for this purpose if what you want to do is compare another your edge detector with the standard one. The "ideal edges" are, as you say, the edges from whatever standard you've chosen.

The distance between an actual and an ideal edge is just the distance in the image, computed using Pythagoras' theorem on their coordinates. This assumes that you have some way of matching up corresponding edges, which may not be trivial.

There is a single summation because you sum across edges, not across pixels.

I'm not sure what pseudocode would add to this.

I recommend looking up Pratt's original paper or book. It's likely to be more understandable and give more details than the very brief summary in CVonline.

[Vivek Bhadouria]

Thanks for the help

Answer 2

[Mohammed Tarek GadAllah]

**Sir, Vivek take care that the site you used for Pratt's FOM http://homepages.inf.ed.ac.uk/rbf/CVonline/LOCAL_COPIES/OWENS/LECT6/node2.html

the formula written in wrong terms

you can see the true formula just for example in the paper sited by:

http://www.matmidia.mat.puc-rio.br/sibgrapi2009/media/posters/59906.pdf

----------------- Really I;m facing the same problem i cant find a code or understand how to calculat The distance between an actual and an ideal edge is just the distance in the image, computed using Pythagoras' theorem on their coordinates

-----------------

if you can help me by a matlab code i'll be happy thanks

Mohammed Tarek

[Mohammed Tarek GadAllah]

You Can take code to help from the following site But, Exactly i don't know it is accurate true Pratt's FOM or not

you can see

the site is :

http://en.pudn.com/downloads201/sourcecode/graph/texture_mapping/detail947301_en.html

Mohammed Tarek

for mailing : [e-mail]