Image regions, which are not easily affected by moving target, are selected as matching region respectively. Then the matching regions are divided into a specified number of sub-regions according to a given experience value. By a presented algorithm of representative point matching(RPM), the local motion vecters(LMV)can be obtained. Median filter and mean filter are applied to get the real jitter vector, with historical motion vectors being considered. Then the video image is compensated according to the jitter vector. The effectiveness of proposed approach is verified based on the DM642 DSP video acquisition and processing experiment platform. The experimental results show that the proposed approach not only has high accuracy of electronic image stabilization(EIS), but also has good real-time performance.
展开▼
