The Feature Point Extraction And Matching Based On The Three-dimensional Imaging Of The Light Field Photography

Image feature extraction and matching is an important component in the applications of computer vision. This paper focuses on the multi-view images matching by the light field camera exposure. Among them, the light field camera imaging principle is recorded the data of beam directions, without complex focus system, simple operation, fast shooting. So it can be widely used in multi-view stereo display and security monitoring and other fields. In this paper, the research is not only the basis of the three-dimensional reconstruction, but also light field camera can achieve the key technology of the refocusing after the fact in digital light filed photography.So far, there are many ways to achieve the image feature matching, including image feature detection, feature description and matching. This paper has the exploration and research on the three aspects, the details are as follows:The first step, image feature detection. The commonly used methods are the methods based on the image edge information, on corner points information and on various interest operators. But the traditional methods has the problems of the large amount of calculation and time-consuming. In order to solve this problem, this paper uses a feature detection based on the gray information of the image--FAST operator(features from accelerated segment test). This method combined with the decision tree theory is effective to improve the speed of feature extraction of images.The second step, feature description. Description of feature point is a local expression by BRIEF(binary robust independent elementary features). Due to the BRIEF descriptor does not have rotational invariance. So in this paper, we first give a direction, then describe the feature point, and generate a binary string.The third step, image matching. There are many ways to measure the similarity. We use the method of nearest neighbor search to find the feature points by the nearest Euclidean distance. It has simple calculation. The experimental results show that the method has the high efficiency, fast speed.The simulation experiments demonstrated that the nearest neighbor search method for image matching based on ORB has high matching accuracy, fast calculation speed. It also has a great advantage to calculate the large number of image arrays by the light field camera.