A Digital Image Stabilization Method Based on the Hilbert–Huang Transform

In this paper, an innovative technique for digital image stabilization (DIS) based on the Hilbert-Huang transform (HHT) is proposed. It exploits the basic features of the HHT in order to separate the local motion signal obtained from an image sequence into two different motion vectors. A variety of embedded systems equipped with a digital image sensor, such as handheld cameras, mobile phones, and robots, can produce image sequences with an observed motion caused by two different types of movements: the smooth camera motion (intentional) and the unwanted shaking motion (jitter). The HHT has been successfully employed in applications for disaggregating signals into smaller portions with specific features (e.g., biomedical applications). Subsequently, for DIS, local motion vectors of an image sequence are calculated, and they are processed by the HHT in order to define both signals. The original signal is divided into a number of waveforms, called intrinsic mode functions (IMFs), using the process of empirical mode decomposition. Hilbert transform is applied to each IMF so that the energy content could be designated. Based on the basic features of the unwanted shaking phenomena (high frequencies and small power contents), intentional and jitter motions are determined, and thus, motion compensation is applied in order to eliminate possible fluctuations and produce an image sequence with smoother transitions. Experimental results have shown that the method not only successfully separates these two different signals but also outperforms other DIS methods.