Corner detection method combining first-order and second-order generalized Gaussian directional derivatives

Corner detection holds a pivotal role in computer vision and image processing domains, as its efficacy directly influences the performance of subsequent tasks like 3D reconstruction, target tracking, and image matching. Our research reveals that current corner detection methodologies do not fully capitalize on first- and second-order structural information across multiple scales and directions, indicating potential for enhanced detection accuracy and localization precision. To address this, we introduce a novel multi-scale, multi-directional corner detection algorithm that integrates first- and second-order generalized Gaussian directional derivative filters, offering a more nuanced distinction between edges and corners. Experimental findings demonstrate that our proposed algorithm surpasses existing corner detection techniques in terms of overall performance.