基于SIFT变换和RBF神经网络的彩色全景图拼接算法

目的 为了复原存在平移、色彩差异、旋转、形变等问题的全景图，提出一种结合SIFT（尺度不变特征变换）和RBF神经网络的彩色全景图拼接算法。方法 通过SIFT算法匹配出两子图中对应的特征点，利用仿射变换解决图像间的旋转和形变问题，采用RBF神经网络纠正子图的色彩差异，最后利用权值矩阵融合技术实现重叠区域的融合。结果 文中算法在拼接效果上优于其他算法，其拼接效果DoEM值为0.902，图像重叠区域过度平滑，有效地避免了融合区域的亮度块或亮度线。结论 该算法效果好，可解决全景图复原过程中多方面的难题。

Color Panorama Mosaic Algorithm Based on SIFT Transform and RBF Neural Network

In order to restore the panorama with the problems of translation, color difference, rotation and deformation, a color panorama stitching algorithm combining SIFT (scale invariant feature transform) and RBF neural network is proposed. The SIFT algorithm was used to match the corresponding feature points in the two subgraphs, and the affine transformation was used to solve the rotation and deformation problems between images. The RBF neural network was used to correct the color difference of the subgraphs. Finally, the weight matrix fusion technique was used to realize the fusion of overlap areas. The proposed algorithm was superior to other algorithms in the stitching effect. The DoEM value of the stitching effect was 0.902, and the image overlap area was excessively smooth, effectively avoiding the luminance block or the luminance line of the merged area. The proposed algorithm has good effect and can solve many problems in the panorama restoration process.