A new image stabilization algorithm based on improved diamond search combined with adaptive partitioning optimization
Abstract
Since the traditional block matching method in the larger search window tend to fall into local optimum to be unsuitable solving large pan shake image stabilization problem, a new image stabilization algorithm is proposed, which is still based on diamond search, but the pattern of large template is modified to ring, search domain is partitioned and sorted by size, global motion vector is extracted through maximum sample statistics and compensated through adaptive mean filter. A numerical example is calculated to show that the method has a higher search efficiency than exhaustive search, three-step search, four-step search and traditional diamond search; maximum sample statistics avoid the prospect's interference effectively when global motion vectors is extracted; adaptive filter compensation has better real-time and smoothness than a fixed size one; PSNR of images after processed is higher significantly than the one before processed. Therefore the results above verify the accuracy of the algorithm and feasibility.
- Publication:
-
2013 International Conference on Optical Instruments and Technology: Optoelectronic Imaging and Processing Technology
- Pub Date:
- December 2013
- DOI:
- 10.1117/12.2038287
- Bibcode:
- 2013SPIE.9045E..1XZ