An image matching method based on locality sensitive confidence evaluation

Abstract

The invention discloses an image matching method based on local sensitivity confidence evaluation. The image matching method comprises the following steps: step 1, local feature extraction and pre-matching; And step 2, searching for a neighbor matching pair. And step 3, calculating a voting line based on neighbor matching and estimating the position of a real matching point. And step 4, defining amatching confidence coefficient and evaluating the matching confidence coefficient. And 5, performing iteration to remove error matching and correct iteration regression matching so as to obtain a final result and display the final result. The method has the advantages that the problems that in an existing method for estimating global transformation between input images or achieving a matching solution through global optimization, abnormal values are likely to appear, interference is likely to happen and the like are solved, and the matching efficiency and accuracy of image matching are effectively improved.

Description

technical field [0001] The invention belongs to the technical fields of image processing, computer vision and multimedia information, and in particular relates to an image matching method based on local sensitive confidence evaluation. Background technique [0002] Establishing reliable feature correspondences plays an important role in various applications of computer vision, including object recognition, common pattern recognition, image retrieval and 3D scene model building. Given a pair of images, an initial correspondence between features is usually obtained by comparing the distance between features extracted using a local feature detector and a descriptor, before locating the correct match from a putative initial matching set. [0003] Existing researchers have made a lot of efforts in this area to locate the real matching pairs from the initial matching set. Nevertheless, the development of algorithms is still difficult to perfectly solve many real-world image probl...

AI Technical Summary

Problems solved by technology

First, the set of putative matches usually contains a large number of false matches, and the situation will be worse if the image is occluded, the object is distorted, etc.

Second, the objects in the general input picture may share a consistent transformation in the local area, rather than the global view, which limits the application of algorithms such as RANSAC

Third, encoding transformations for local regions, while applying them to a global optimization framework for correct matching recognition, often has a large amount of noise

Images