Skip to Main Content
This study is motivated by the emergence of a new class of tunable infrared spectral-imaging sensors that offer the ability to dynamically vary the sensor's intrinsic spectral response from frame to frame in an electronically controlled fashion. A manifestation of this is when a sequence of dissimilar spectral responses is periodically realized, whereby in every period of acquired imagery, each frame is associated with a distinct spectral band. Traditional scene-based global shift estimation algorithms are not applicable to such spectrally heterogeneous video sequences, as a pixel value may change from frame to frame as a result of both global motion and varying spectral response. In this paper, a novel algorithm is proposed and examined to fuse a series of coarse global shift estimates between periodically sampled pairs of nonadjacent frames to estimate motion between consecutive frames; each pair corresponds to two nonadjacent frames of the same spectral band. The proposed algorithm outperforms three alternative methods, with the average error being one half of that obtained by using an equal weights version of the proposed algorithm, one-fourth of that obtained by using a simple linear interpolation method, and one-twentieth of that obtained by using a naiÂ¿ve correlation-based direct method.