Fast digital image stabilization using constrained one-bit transform, sparse search points, and partial distortion search

Several improvements are proposed to speed-up the two dimensional full search based digital image sequence stabilization process in this work. Firstly, the kernel used in the constrained one-bit transform based stabilization approach is modified to reduce the computational load. Next, the utilization of a sparse search point approach is investigated. Finally, a partial distortion search approach using spiral scanning combined with a sparse search point approach. Experiments show 10 to 20 times reduction of computational complexity without much decrease in the performance.