The Community for Technology Leaders
Green Image
<p>Edges in a scene generally project to smooth continuous curves nearly everywhere in the image, which results in low angular deviation of the intensity gradients in small neighborhoods straddling edges. Angular deviation is shown to be a measure of SNR. A theoretical analysis of angular deviation arising due to independent and identically distributed N(0, sigma /sup 2/) random noise is presented. Angular deviation thresholds for neighborhood sizes from 3*3 to 11*11 pixels are determined both from this analysis and numerical examples. The proposed gradient angular dispersion detection algorithm detects edge elements by comparing the measured angular deviation with values computed for the minimum acceptable SNR. Low values of deviation violate the 'no edge' hypothesis. The algorithm is shown to make good use of the limited dynamic range of the imaging system. The sensitivity and selectivity of the strategy are both shown to be high.</p>
image recognition; edge detection; angular deviation; distributed random noise; angular dispersion; gradient direction; intensity gradients; gradient angular dispersion detection; sensitivity; selectivity; 11 pixels; 121 pixels; 3 pixels; 9 pixels; edge detection; noise; numerical analysis
P.H. Gregson, "Using Angular Dispersion of Gradient Direction for Detecting Edge Ribbons", IEEE Transactions on Pattern Analysis & Machine Intelligence, vol. 15, no. , pp. 682-696, July 1993, doi:10.1109/34.221169
86 ms
(Ver 3.3 (11022016))