拉普拉斯特征耦合方差度量的图像匹配算法

为了解决当前图像匹配算法主要是通过像素点间的距离信息来实现特征匹配，忽略了图像间的方差信息，导致匹配结果中存在较多的错误匹配等不足，本文提出了一种基于拉普拉斯特征制约与方差度量的图像匹配方法。首先，引入Harris算子，对图像特征进行粗提取，并利用像素点的拉普拉斯特征，删除伪特征点，对粗提取的图像特征进行优化，获取更为准确的图像特征。然后，依据图像的梯度特征来计算图像特征的方向信息，以此建立特征点的邻域，通过求取该范围内的Haar小波值，从而得到特征向量。采用区域方差模型对图像的方差信息实施度量，并联合特征点的欧氏距离，对特征点进行更为准确的匹配。最后，采用随机样本一致性(RANSAC)机制对特征匹配结果实施优化，剔除其中的错误匹配，从而完成图像匹配。实验结果显示：较当前较为先进的匹配算法而言，在旋转、缩放等几何变换干扰下，所提算法具备更高的匹配准确率，维持在90%以上。

Image matching method based on Laplacian feature coupling variance measure

Current image matching algorithms mainly use the distance information between pixels to achieve feature matching, ignoring the variance information between images, resulting in more false matching in the matching results. An image matching method is proposed based on Laplacian feature constrained coupling variance measure. Firstly, Harris operator is introduced to extract image features roughly. On the basis of rough extraction, Laplacian feature of pixels is utilized to optimize the extracted image features in order to obtain more accurate image features. Then, the gradient feature of the image is employed to calculate the direction information of the image. Based on the gradient feature, the neighborhood of the feature points is established, and the Haar wavelet value in the neighborhood is solved to obtain the feature vector. Finally, the regional variance model is adopted to measure the variance information of the image, and it is introduced into the process of image feature matching. The variance information is added on the basis of Euclidean distance measurement of feature points to achieve image feature matching more accurately. Random Sample Consensus(RANSAC) method is adopted to purify the results of feature matching, eliminate mismatching and complete image matching. The experimental results show that compared with the existing matching algorithms, the proposed algorithm has better matching performance and higher accuracy, with accuracy above 90%.