运动目标在图像拼接时, 易产生重影现象。在对SURF算法进行改善的基础上, 提出了自适应去除重影的拼接算法。利用增强的单通道图像对其特征点进行提取、匹配, 再利用帧差法确定出运动目标在图像重叠区域的位置; 取运动目标区域的中心点作为参考点, 采用改进的加权融合算法求取加权系数, 以增大权值差异。实验表明所提方法能较好地解决运动目标对图像拼接的干扰问题, 与最优拼接缝多频率融合算法相比,融合质量、清晰度以及拼接的实时性得到了提高。

The technique of image mosaic may cause ghost phenomenon when it processes the image with moving objects. An adaptive algorithm was proposed to remove ghosting based on optimizing previous SURF algorithm. The enhanced image was used to replace the original gray image processed by SURF, thus the feature points were extracted. Then, it locates the moving objects in the overlap region of two images using frame differential method. And the improved weighted fusion algorithm was used to calculate the weighted coefficient so as to increase the weight difference. Experimental results show that this algorithm can solve the interference problem produced by moving objects, thus compared to the algorithm of using the optimal seam based on multi-frequency fusion, the proposed method can obtain better fusion quality and highher image definition.