Abstract In detecting patterns in data arranged in a spatial field of one, two or higher dimension, a pattern contrast signal is evaluated which measures the similarity to the pattern of the data around a location in the spatial field. The pattern is subsequently detected if the pattern contrast signal is sufficiently strong. There is, however, a risk of erroneous detection when a coincidental noise configuration causes an increased pattern contrast signal. The invention provides for a method and apparatus to discriminate against such erroneous detections. For this purpose, the method observes the extent to which a would-be pattern dominates the effect on the pattern contrast signal of the distribution of data values in the surroundings of the location, this distribution being taken to be typical of noise. In one embodiment, this is done by confirming that the pattern is sufficiently strong to suppress any other would-be pattern detections in the surroundings. In another embodiment, this is done by confirming that the pattern contrast signal responds as expected in the noise free case when the spatial field is rescaled around the location before evaluating the pattern contrast signal.
A method for distinguishing noise responses from true pattern responses comprising the steps of a) receiving an input signal; b) using first and second pattern detectors to generate first and second pattern contrast signals, the first and second pattern detectors having respective first and second edge extraction operators operating on input areas, the input areas having different input widths; c) comparing the first and second pattern contrast signals at a candidate location; and d) distinguishing the candidate location as containing a true pattern response based on a result of said comparing step.
Claim 2
Independent Claim
92 words
A method for distinguishing noise responses from true pattern responses comprising the steps of a) receiving an input signal; b) using first and second pattern detectors to generate first and second pattern contrast signals, the first and second pattern detectors having respective first and second edge extraction operators operating on input areas, the input areas having different input widths; c) comparing the first and second pattern contrast signals at a candidate location; and d) distinguishing the candidate location as containing a true pattern response based on a result of said comparing step.
Claim 3
Dependent Claim
39 words
The method of claim 1 comprising the steps of sharpening said first pattern contrast signal to form a sharpened pattern contrast signal; and using the sharpened pattern contrast signal as the first pattern contrast signal in said comparing step.
Claim 4
Dependent Claim
27 words
The method of claim 1 comprising the step of computing said second pattern contrast signal for candidate locations having non-zero responses in the first pattern contrast signal.
Claim 5
Dependent Claim
24 words
The method of claim 1 wherein said comparing step comprises generating a pattern ratio signal; and comparing the ratio signal with a predetermined criterion.
Claim 6
Dependent Claim
30 words
The method of claim 1 wherein said comparing step comprises comparing said first pattern contrast signal with values computed by multiplying said second pattern contrast signal by two different coefficients.