Research On Multi-Modality Image Matching Technology Based On Gross Contour

Image matching technology is widely used in the fields of pattern recognition, navigation and guidance, medical diagnosis, as well as in computer vision, especially in the military field. This paper studies gross contour-based multi-modality image matching technology on the basis of human's cognition process of image.First, the formation of gross edge was researched according to the imaging principles and characteristics of multi-modality image. Then the definition of gross edge was presented and a multi-modality image matching technology based on gross contour was proposed in this paper.Second, the gross edge was captured by gross edge detection algorithm. According to the imaging principles and the gray distributions of multi-modality image, a novel edge detection algorithm based on force field transformation was proposed. The force field magnitude and direction of each pixel could be calculated by the concept of gravitational force. By analyzing the properties of force magnitude and direction at gross region, the gross edge detection method was proposed accordingly. Then the gross contour of image was captured by contour tracking.Third, in order to eliminate the impact of noise and scale variation, the contour fitting methods based on Douglas-Peucker and curvature scale-space were researched, then the relationship of scale variation and image resolution was obtained through experimental analysis, and the same scale smooth contours were captured.Finally, contour matching algorithm was researched. By analyzing the current methods of contour curve representation and contour matching, a novel closed contour curve representation based on reciprocal integration of centroid distance and a new open contour curve representation based on local curvature scale were proposed. Then the match points of contours were captured through measuring the similarity of invariance of contours, and the multi-modality image matching based on gross contour was realized. Experimental results show that the proposed technology eliminates the impacts of noise and scale variation effectively, and it has better robustness to multi-modality image matching which has the characters of rotation, scale and translation variation.