Homogenized ORB Algorithm Using Dynamic Threshold and Improved Quadtree

The Oriented FAST and Rotated BRIEF (ORB) algorithm has the problem that extracted feature points are overconcentrated or even overlapped, leading to information loss of local image features. A homogenized ORB using dynamic thresholds improved quadtree method is proposed in this paper, named Quadtree (QTORB). In point extraction stage, a new threshold calculation enhance algorithm’s ability extract at homogeneous regions. Then, adopted manage optimize eliminate those excessively concentrated overlapping points. Meanwhile, optimization process, different depths set pyramid levels prevent excessive splitting increase speed. description gray difference value introduced saliency description. Finally, Hamming distance used match RANSAC avoid mismatches. Two datasets, namely, optical dataset SAR dataset, experiment. experimental result shows that, considering accuracy real-time efficiency, QTORB can effectively improve distribution uniformity