为了将同一场景中具有重叠区域的序列图像合成一幅宽视角、高分辨率的图像, 提出了一种结合尺度不变特征变换(SIFT)算法和Canny特征边缘提取的方法进行图像拼接。实验结果表明所提方法与SIFT算法相比, 其参考特征点及待拼接特征点所需个数分别降低了26.56%和51.71%, 匹配点对数降低51.65%, 图像拼接用时降低了9.49%。此方法提高了图像拼接的精确性和实时性, 能较好地解决图像间存在光照、旋转、尺度变换、仿射等问题, 实现无人工干预的自动拼接。

In order to compound the sequence images with overlapped areas in the same scene, such as virtual scene reconstruction, medical and satellite detection, to be images with wide viewing angle and high resolution, a method combining Scale Invariant Feature Transform (SIFT) and Canny edge feature extraction was proposed. Experimental results show that, compared with the SIFT algorithm, the proposed method enables the number of feature points in the reference image and the mosaic image be decreased by 26.56% and 51.71%, respectively, the number of matching features also decreases by 51.65%, and the mosaic speed of generating panorama decreases by 9.49%. This method improves the accuracy and real-time of image mosaic, and can solve the problems such as rotation, scale transformation and affine among the images.