基于邻域梯度图像拼接算法研究  

Research of Image Mosaic Algorithm Based on Neighborhood Gradient

作　　者：张芳芳[1] 

机构地区：[1]河南城建学院,河南平顶山467036

出　　处：《科技通报》2012年第10期61-63,共3页Bulletin of Science and Technology

摘　　要：图像配准是图像拼接过程中至关重要的一步,本文首先分析了SIFT(scale invariant feature trans-form)算法原理,然后根据SIFT算法的缺点对其进行改进。通过Harris角点检测算法,剔除不稳定的边缘极值点,从而减少了特征点的数量;同时采用RANSAC算法去除伪匹配点。实验结果表明,改进算法不仅具有旋转、尺度变化不变性,对于噪声和亮度变化具有良好的鲁棒性,而且匹配速度比改进前大大提高。Image registration is a crutial step in the process of image mosaic.Scale Invariant Feature Transform algorithm is analysised in this article,and an improved SIFT algorithm is worked out according to the disadvantage of the original SIFT.Edge extreme points is removed by means of Harris corner detection algorithm,which reduce the number of feature points.RANSAC algorithm is also used to remove false match points.The result of experiment shows that the improved algorithm not only obtain rotation and scale invariance but also has good robustness to noise and brightness change.It can improve matching speed greatly than ever.

关 键 词：尺度不变特性变换 特征提取 特征匹配 邻域梯度 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]