Search

Highlights • Propose a method of combining lines and keypoints and matching them on multispectral images. • Design a line descriptor that does not require ending points to be accurately detected. • The proposed method is applicable to multispectral image on which corresponding lines are not integrally detected. Abstract In this work, we proposed a method of combining lines and keypoints and matching them on multispectral images. Existing keypoint techniques usually assume images contain abundant texture information and employ hand-crafted gradient information to match keypoints. Unfortunately, mutltispectral images may lack sufficient texture information and gradient information is unable to represent well the common feature, which makes the matching of keypoints very unreliable. Lines on the contrary can be well extracted and serve as a complementary and dual image featues, but robustly describing neighboring region around lines is difficult due to inaccurate ending points and line fragmentation. In observance of the strength of lines and keypoints,this work proposes combining the keypoints and lines and matching them together. Firstly,keypoints and lines are extracted and grouped according to spatial affinity. Then we sample points in the local region for each combination and the sampled points are then used to build descriptors for each group which will be used to match the combinations of keypoints and lines. The main contribution is that the proposed method can achieve reliable feature matchings without requiring accurate end points oweing to the use of combination of keypoints and lines. The proposed method is tested on a large number of multispectral images and experimental results show that it can effectively match lines and keypoints simultaneously. [ABSTRACT FROM AUTHOR]

Due to non-linear intensity changes between multispectral images, the existed descriptors often yield low matching performance. In order to build reliable keypoint mappings on multispectral images, a novel coarse-to-fine method is designed using projective transformation and the information of edge overlap. The method consists of a coarse process and a fine-tuning process. In the coarse process, initial keypoint mappings are built with the descriptors associated with keypoints and the relative distance constraints are employed on them to remove outliers. In the fine-tuning process, the edge overlap information is utilized as similarity metric and an iterative framework is applied to search correct keypoint mappings. The performance of the proposed is investigated with keypoints extracted by speeded-up robust features. The experiment results show that the proposed method can build more reliable keypoint mappings on multispectral images than existed methods. [ABSTRACT FROM AUTHOR]

This paper proposes an approach to registering multispectral images by establishing keypoint matches. The matching ability of descriptors is characterized by the repeatability and distinctiveness that typically decrease on multispectral images. The decrease of matching ability often yields a set of keypoint matches containing a high rate of incorrect matches, and in this case the outlier matches are very difficult to be removed. To establish reliable keypoint matches, this paper proposes an approach of two stages. Firstly, keypoint matches of smaller descriptor distance are obtained as an initial set. Secondly, complementary information to the local window for computing descriptors is employed to evaluate keypoint matches and find good matches. A smaller descriptor distance for a keypoint match implies a greater probability of being correct and hence the initial set contains a higher rate of correct matches. The global information can be viewed as a means of enhancing the matching ability of descriptors, compensating the decrease of common information between multispectral images. Experimental results show that the proposed method can effectively establish keypoint matches on multispectral images of large spectral difference. [ABSTRACT FROM AUTHOR]