Image processing

Hello everyone...... I have a problem to find intersection point of two curves...If I have FAR(False Acceptance Rate) for 20 threshold values and FRR(False Rejection Rate) for those threshold values also....Now my question is how to find out the intersection point of these two curves that denotes EER(Equal Error Rate)...Please reply soon......Advance Thanks........

Answer 1

[Image Analyst]

How about something like

tolerance = .01; % or whatever...
crossingIndex = find(abs(FAR - FRR) < tolerance, 1, 'first');

Is that good enough?

[harjan]

But if I have a set of FAR values for 10 thresholds such as FAR=[200 0.0091; 210 0.0082;......];

and FRR=[200 0.0013;210 0.0023;.....]; like that.If i plot graph for these values it will cut at some point...At that point i want to know the x axis and y axis values....

[Image Analyst]

Yes, that's what I thought. Did you not like my code? If you want more accuracy, though I don't know why you'd need it with only 20 observations, then you can do a regression with polyfit though you may be getting a fake increase in accuracy since the fitted value is not necessarily more accurate. I mean, how accurate can you get to the "true" but unknown value when you have only 20 observations?

Answer 2

[Stephen]

yeah, use Euler's method and just subtract one from another and search the zeros or, as the guy above said, add a tolerance level. min(abs(FAR - FRR)) would work if you didn't care how far off from 0 their difference is.

[harjan]

But if I have a set of FAR values for 10 thresholds such as FAR=[200 0.0091; 210 0.0082;......];

and FRR=[200 0.0013;210 0.0023;.....]; like that.If i plot graph for these values it will cut at some point...At that point i want to know the x axis and y axis values....