Research On Feature Matching Algorithm Of Light Field Image Based On EPI Angle Consistency

Image feature detection and matching is an important foundation for many high-level computer vision tasks.Real scene targets generally have occlusion,non-Lambertian surfaces,complex textures and other factors.The detection and matching of traditional image features still face great challenges due to these factors.Compared with the traditional imaging system,Light field cameras simultaneously capture multiple images of the same scene from different viewpoints in a regular and dense sampling The multi angle information of light field data provides a more advantageous data base than traditional image for high-precision feature detection and matching.Epipolar plane image(EPI)of light field is a two-dimensional slice of four-dimensional light field.Its advantage is that it not only contains the position information and angle information of light field,but also can be calculated by traditional image processing methods.Therefore,this thesis proposes a light field image features detection and matching method based on EPI angle consistency.(1)Based on the traditional image local invariance feature theory,this paper analyzes the local invariance of the biplane light field in space-angle sampling.Aiming at the high-dimensional characteristics of the light field data,this paper uses the angle consistency of the light field in the EPI space,and proposes the correlation of the angle domain characteristics of the light field.(2)According to the angle consistency of EPI,this thesis proposes a light field image feature point detection method based on EPI line detection.In this method,the line detection is performed on the EPI image of light field,and then the line detection results are projected from the EPI space to the sub aperture image space of light field to realize the key point detection of light field image.Experiments prove that the method proposed in this article has better performance.(3)Based on the EPI line detection method of light field image feature point detection,a feature detection and matching algorithm based on the angle consistency of the light field EPI is proposed.This method performs SURF feature detection in the light field angle-space domain,and uses the angle consistency of the light field to the polar plane image as a constraint to screen the candidate feature matching.Experiments show that the proposed method can eliminate a large number of o cclusion and non Lambert surface factors caused by mismatching,and effectively improve the accuracy of light field image features matching.(4)In order to verify the features detection method of light field image based on EPI line detection and the feature detection and matching algorithm based on the angle consistency of light field EPI proposed in this thesis,this thesis constructs light field image matching datasets of different scenes.In different datasets,the experimental results prove the effectiveness and accuracy of the method proposed in this paper.In summary,this thesis proposes a light field image features detection and matching method based on EPI angle consistency,and constructs light field feature matching datasets for different scenes.In different datasets,the experimental results prove that the special light field image detection method based on EPI line detection proposed in this thesis angle consistency has a higher accuracy rate.