Digital image stabilization method based on variational mode decomposition and sampling fluctuation analysis

Unintentional motions often cause cameras to produce shaky images, which is a significant source of inter-frame blur and video quality decline. To ad-dress this issue, we present a digital image stabilization approach based on variational mode decomposition (VMD) and sampling fluctuation analysis (SFA) to generate stable video sequences. Our method first estimates the global motion vector (GMV) from a video sequence using the speeded up robust features (SURF) algorithm. We then decompose the GMV into various modes using VMD to separate jitter motions from intentional ones. Here, SFA is applied to distinguish different modes based on their unique structural characteristics. We evaluate our proposed method in complex scenarios by comparing it with several existing methods. Our experimental results demonstrate that VMD outperforms other stabilization techniques under comparable conditions.