INVESTIGATION OF METHODS OF AUTOMATIC STITCHING OF PANORAMIC IMAGES

Abstract

The relevance of panoramic stitching is explained by the fact that powerful computers and image processing algorithms are currently available, which allow you to automatically stitch many images into a panorama with a high degree of accuracy and quality. This makes panoramic stitching an important tool for both professional photographers and amateur photographers, as well as in many other areas related to image processing and computer vision. The leading trend in the development of panoramic stitching is to improve the accuracy and speed of algorithms, as well as to expand the possibilities for working with large amounts of data. One of the directions of its development is the development of tools for creating interactive panoramic images and virtual tours. The paper proposes a method of absolutely automatic stitching of panoramic images using methods of invariant local functions for finding key points and their descriptors, projective transformation using the RANSAC algorithm, image alignment based on the calculation of homographic parameters of the camera, multi-band image mixing. To test the proposed method, a software prototype was implemented, photographs from the Huns exhibition at the M.N. Khangalov Museum of the History of the Republic of Buryatia were taken as experimental data. The results of the software prototype are panoramic images obtained based on the processing of these photos. The conducted computational experiments allow us to conclude that the results obtained show high accuracy when compared with the real world.