Research Of Image Matching And Fusion Algorithm

Image matching and image fusion are two important parts in the field of image processing. Image matching can be described as the search for common information from two different images to perform registration or localization. Image fusion indicates the processing for the image information generated by different sensors,and generate image with richer content. With the development of liquid crystal display technology, virtual meters are widely used in aircraft cockpit, automotive, military and other important fields. Whether in the virtual meters development process or monitoring process, the automatic detection and identification for the virtual meters can save development costs and speed up the pace of automation,therefore it has a very important practical significance.Image matching method of this thesis mainly focuses on the virtual meters image,and the purpose is to quickly and accurately classify and identify the meters image from the image of a variety of meters. This research lays the foundation for the auxiliary development, as well as the automated auxiliary test of the system design process of graphic class. The research of image fusion algorithm mainly based on the infrared image and visible image processing, in order to eliminate the blind area caused by too dark light, to provide a technical basis for night assisted driving and environmental detection and reconnaissance. The main contents of this thesis are as follows:1. This thesis studies the matching criterion and search strategy of template matching in image matching. In view of the shortcomings of time-consuming long and slow speed of template matching, this thesis proposes an improved template matching method based on dual-threshold binary image matching criteria and Pyramid combined with the effective area search. After testing, the improved algorithm reduces the total amount of data involved in the image matching. Optimized search strategy reduces the number of calculations of data. And it has a certain ability to resist the noise, which guarantees the accuracy,and is better than the FFT acceleration method on computing speed2. In this thesis, we study the classification and characteristics of point feature in image matching method, as well as the associated with the matching description method, extraction method. In view of the problems during the extraction process of FAST algorithm with the corners uneven density distribution, poor location and matching process time-consuming, this thesis proposes a FAST joint vertical gradient filter and regions screening of corner extraction improved method. After testing, the improved algorithm makes the corner distribution more uniform, and distribution in the calibration section of the fixed, cuts out the distribution of changes on the pointer points, reduces the total number of corners to participate in the match, improves matching real-time.3. This thesis analyzes the infrared image and visible image fusion method based on spatial domain and frequency domain, as well as analysis and presentation of the results of the evaluation method of fusion. Aiming at the shortcomings of the current algorithm for the fusion of details, this chapter proposes the improved Laplacian fusion algorithm based on the combination of gray enhancement and target area detection. After testing, the algorithm based on the realization of the target area, is more prominent display detail, from both subjective and objective evaluation methods are superior contrast.4. In this thesis, the algorithm are realized by programming, respectively, using the Opencv and MFC design integrate the various modules of the test tool software.