Efficient Iris Recognition Based on Optimal Subfeature Selection and Weighted Subregion Fusion
Figure 5
The process of computation of keypoint descriptor. The corresponding gradient magnitude and orientation at each sample point in a region around keypoint location are computed firstly, as shown in (a). Then these samples are accumulated into orientation histograms summarizing the contents over subregions, with the length of each arrow corresponding to the sum of the gradient magnitudes near that direction within the region, as shown in (b). (a) Image gradients; (b) keypoint descriptor.
