Remote Sensing Image Mosaic Technology Based on Improved SIFT Algorithm

Abstract

To solve the problem of a large amount of computation and low efficiency of the traditional remote sensing image mosaic algorithm, an Improved SIFT feature matching image mosaic algorithm is proposed. Aiming at the disadvantage of too many dimensions of feature point descriptors in SIFT algorithm, a dimension reduction scheme is proposed. Centered on the feature point, concentric circles are constructed with the radius of 2 pixels, 4 pixels and 6 pixels respectively. The region is divided according to the distance from the feature point. If the distance is close, the weight is high, and if the distance is far, the weight is low. The region is divided into 4, 8, and 16 partitions, so as to generate a 56-dimensional descriptor. Furthermore, the feature points in the image matching are divided into two data sets, namely, maximum and minimum. The best matching point is selected by calculating the distance between the maximum and minimum, which reduces the amount of calculation of image matching and saves the algorithm time. Finally, the weighted smoothing method is used for image mosaic. Experimental tests were carried out on the three situations of remote sensing image scaling, rotation, and gray change. The results show that the proposed algorithm can effectively reduce the calculation amount and accelerate the stitching speed, and the stitching effect is ideal.