基于兴趣点检测和仿射变换模型的序列图像拼接

针对序列图像的拼接问题，提出了一种基于兴趣点检测和仿射变换模型的图像配准算法，利用Harris算子检测兴趣点，并引入灰度交叉匹配和兴趣点坐标离散度两级约束策略确定同名兴趣点，通过在图像间建立仿射变换模型实现图像的配准；提出了一种自适应加权归一化函数对拼接图像进行处理，消除由于图像间灰度差异而导致的拼接缝。实验结果表明算法具有较高的配准精度和良好的像素融合效果。

Sequential Images Mosaic Based on Interest-point Detection and Affine Transform Model

For the problem of the sequential image mosaic,a new method for image registration based on the interest-point detection and affine transform model is proposed firstly,which detects interest-points by the Harris operator,and introduces a two-level restricted tactic based on cross match and coordinate scattering to gain the homonymy points.Then image registration is carried out according to the affine transform model.A normalized adaptive weighted function is given for eliminating the mosaic-sew caused by the difference of the pixel gray in the images,which makes the smoothly transition of the mosaic image.The experiment results indicate that approach proposed has good registration precision and nice effect of pixel-fusion.