Research of Descriptor-Based Image Normalization and Comparative Analysis of SURF, SIFT, BRISK, ORB, KAZE, AKAZE Descriptors
In cooperation with experts from the Informatics Department of KNURE, we have researched the descriptor-based approach to image normalization. In experiments, we've used key points and SURF, SIFT, BRISK, ORB, KAZE, AKAZE descriptors to find out the normalization parameters. As a result, we've obtained a comparison of the normalization process's quality and time costs based on different descriptors.
The research outcomes may help solve computer vision's tasks where it is necessary to make decisions on local features of images, like detecting, tracking, tagging, or recognizing image plagiarism. We've studied each step of normalization in detail, developed a mathematical and software model for further studies on the photo image pairs normalization with the different descriptors, created the dataset for experiments and drawn conclusions and takeovers for further implementation.