针对图像镶嵌过程中如果图像序列中存在运动目标就会引起重影的问题,文中提出了一种新的去除运动目标重影的图像镶嵌方法。算法首先对视频图像进行了运动分割,在图像匹配阶段采用边缘特征点进行匹配,并由马尔可夫随机场模型生成运动目标的二值模板,剔除掉运动目标二值模板上的边缘点,从而保证图像匹配的准确率。在镶嵌阶段使用活动轮廓模型生成一条最优镶嵌线,产生的镶嵌线充分考虑了图像的边缘特征和梯度信息,确保了镶嵌后图像两边纹理差异更小。从连续视频帧中选取多帧图像进行实际的图像镶嵌,实验结果表明文中算法取得了较好的效果。
For the problem of moving targets occurring in image sequences will cause ghost phenomenon,propose a new image mosaic method to remove the ghost of moving targets. At first,the algorithm takes motion segmentation to the video images,and uses edge fea-ture points for image matching,then generates a binary template by Markov random field,and eliminates the feature points which exist on the template,thus ensuring the accuracy of image matching. In the blending phase,use active contour model to generate an optimal mosaic line,the resulting mosaic line take full account of the image edge features and gradient information,to ensure that the texture differences of both sides of the fused image smaller. Selecting multi-frame images from successive video for the image mosaic,and the experimental results show that this algorithm has achieved a better image mosaic effect.