Image feature matching method for multi-scale detection based on anisotropic diffusion operation

Abstract

The invention relates to an anisotropic diffusion operation-based multi-scale detection image feature matching method, which comprises the following specific steps of: introducing anisotropic diffusion to filter an image to be processed, and then solving a nonlinear diffusion filter equation by using a mathematical framework rapid explicit diffusion scheme to obtain an anisotropic diffusion image; a nonlinear scale space is established; after the scale space is successfully established, entering a feature detection stage, calculating a response value of each point in the image through a Hessian matrix, and detecting a local maximum point in the image according to the response value as a feature point; the feature points are endowed with main directions, descriptor generation operation is carried out by using a specific algorithm, and then image feature matching is carried out by using descriptor similarity measurement; the method is applied to a natural feature-based tracking registration technology, and has a strong practical value in an environment with relatively high requirements on real-time performance and effect.

Description

technical field [0001] The invention relates to a method of constructing a nonlinear scale space based on an anisotropic diffusion method to detect feature points to realize image feature matching, especially an image feature matching method based on multi-scale detection of an anisotropic diffusion operation, belonging to The technical field of computer vision graphics and image processing. Background technique [0002] Graphics and image processing technology is one of the important research contents of computer graphics, and it is also an important part of many computer vision applications such as virtual reality, augmented reality, 3D reconstruction, target recognition, and target tracking. The purpose of image feature matching is to find the homogeneous regions between images, and then establish the corresponding relationship between images according to the mapping of homogeneous regions. However, due to the influence of illumination changes, viewing angle changes, tar...

AI Technical Summary

Problems solved by technology

However, due to the influence of illumination changes, viewing angle changes, target shape changes, image noise, etc., the appearance of the image to be matched may change to varying degrees, thus affecting the matching effect.

Images