An Improved Image Registration Method of Minority Costume Based on K-means

The matching of minority costume images is not only the field of image registration but also difference from the general image registration. In the image of minority costumes, there are cases where the color is rich and the redundancy is more complicated, and there are a lot of bright colors and many of the same features. Based on these characteristics of minority costume images, we propose K-Means cluster preprocessing for images in the registration of minority costumes, so as to avoid the influence of some redundant points or noises in subsequent registration and thus improve the accuracy of registration. Then we use the classic registration algorithm to register the minority costume images and finally achieve the registration of minority costume images from two different angles.