【摘要】 提出了一种新的基于轮廓特征的图像拼接算法。在特征提取阶段 ,对卷积图进行增强 ,并用区域增长方法进行辅助校正 ,能改善轮廓提取效果 ;在特征表示方面 ,用形状签名代替链码来描述轮廓 ,从而提高了计算速度 ,降低了由噪声干扰和镜头形变造成的影响。实验结果表明 ,该算法在速度上优于基于链码的配准算法 ,准确率高 ,并且可以适用于未经过几何形变校正的图像的配准与拼接。更多还原

【Abstract】 A new feature based approach to automatal image to image mosaicing was presented. In the feature extraction process, the performance of the Laplacian of Guassian operators was improved after enhancing the resultant convolved image and calibrating with region growing algorithm. The chain code was replaced by the shape signature and used to express the contour, improve the speed and decrease the effect of noise and transformation due to camera calibration. Experimental results verified the better robustness and efficiency of the algorithm than chain code based algorithm. It can be used to image mosaicing without camera calibration.更多还原