基于改进的快速Shi-Tomasi特征点检测算法的图像拼接算法

Image Mosaic Algorithm Based on Improved Fast Shi-Tomasi Feature Points Detection Algorithm

针对Shi-Tomasi算法计算量大导致的特征点提取速度慢的问题,利用FAST算法对Shi-Tomasi算法进行加速,提出基于此改进算法的快速图像拼接算法。该算法首先应用FAST算法对整幅图像进行处理,基于筛选点邻域再采用Shi-Tomasi算法获取特征点,接着用NCC法进行特征点匹配,然后用RANSAC法剔除错误匹配并完成图像配准,最后用渐入渐出法进行图像融合。对比实验结果表明,改进算法相对基于ShiTomasi的图像拼接算法速度大幅提升,特征点匹配的正确率有所提高,拼接完成后图像的视觉效果也得到保持。

Since the speed of Shi-Tomasi feature points detection algorithm is slow because of complicated calculation,an image mosaic algorithm based on an improved Shi-Tomasi algorithm accelerated with FAST algorithm is proposed. This algorithm first processes the whole image with FAST algorithm. Then neighborhoods of selected points are processed with Shi-Tomasi algorithm to obtain ultimate feature points. Next,NCC algorithm is used to match feature points,and RANSAC algorithm is used to remove mismatches and achieve image registration. At last,image fusion is achieved by weighting algorithm. Experiment result shows that relative to image mosaic algorithm based on Shi-Tomasi algorithm,improved image mosaic algorithm is significantly faster,its accuracy of feature point matching is higher,and visual effect of ultimate image which is stitched with this algorithm can be maintained.