基于自适应模板匹配的帧间全局运动估计算法

Interframe Global Motion Estimation Algorithm Based on Adaptive Template Matching

摘要:

针对传统的匹配方法在匹配模板与待匹配图像间存在噪声影响、亮度等差异时导致匹配算法在时间和精度上得不到很好的统一, 由此影响到红外图像帧间全局运动估计和补偿问题, 提出了基于自适应模板匹配的方法进行帧间全局运动估计算法。该算法首先利用模板选择策略进行待匹配模板的选取, 提高匹配的精度;然后提出自适应模板匹配准则, 以达到较好的匹配效果, 克服噪声等奇异点对误差函数值的影响;最后提出菱形搜索策略, 以便搜索到最佳匹配点, 使搜索不至于陷入局部最优, 并提高了搜索速度。仿真实验结果表明, 在红外图像背景变化较为缓慢的情况下, 所提算法降低了帧间全局运动估计计算复杂度, 同时具有很好的匹配精度和准确性。

Abstract:

For traditional template matching methods, differences such as noise effect and brightness will cause arithmetic asynchrony on time and precision in matching algorithm. Thus the estimation and compensation of the global motion between frames of infrared images will be affected, in view of this, an inter-frame global motion estimation algorithm based on adaptive template matching is proposed. Firstly, it uses the template selection strategy to select the template to be matched so as to improve the matching precision; Then, an adaptive template matching criterion is put forward to achieve good matching effects, so as to overcome the effects of the singular points such as noise on the value of an error function; Finally, the diamond search strategy is put forward to search the best matching point so as to stop the search into local optimum, thus the search speed can be improved. Simulation results show that if the background of the infrared images changes much slowly, the proposed algorithm can reduce the computational complexity and ensure good matching precision and accuracy.