Temporally forward nonlinear scale space with octave prediction for high frame rate and ultra-low delay a-kaze matching system

High frame rate and ultra-low delay matching system is appealing because of its excellent experience for human-machine interactive applications. A-KAZE algorithm is chosen because of its high robustness and high speed. Nonlinear scale space is very important in A-KAZE, but it not only has at least one frame delay and but also is not hardware friendly. This paper proposes temporally forward nonlinear scale space with octave prediction for high frame rate and ultra-low delay A-KAZE matching system. Problems of complex calculations, data dependency and long time delay are solved by HFD based temporally forward nonlinear scale space with octave prediction. Motion estimation prediction is utilized to improve the robustness. It is also processed parallel with keypoint detection part. It finishes processing before the next frame coming and there is no delay as a results. What's more, lower position gray-coded bit-plane motion estimation has been proposed to improve performance. The results show that the proposed method keeps F-score more than 95% for most cases and shows much better performance compared with the current high frame rate and ultra-low delay matching system.