Research And Application Of Diffusion Morphologic Method

With the development of image processing theory and technology,image processing is applied more and more widely in the fields of industry,medicine,computer vision and so on.The requirement for algorithm stability in the research and development of image processing technology is more and more higher.The image recognition technology and image matching technology has become an important technology has received extensive attention in image processing among them,the image similarity research is one of the most important research problems of recognition and matching technology,there have been many mature research methods and results can be summarized into the related research methods of the image to the shape,and the shape to shape.At present,in the research of image to shape similarity,the research results and algorithms based on the Pyt'ev morphology analysis theory have been concerned.The Mosaic image and its related algorithm depend on the initial segmentation of the image,so the algorithm has some defects in stability.To solve this problem,Vizilter applies the ideas of data reduction and diffusion mapping to image similarity research,the related methods to improve the stability.In this paper,based on the research of Vizilter,the method is extended and the related algorithms are improved to open the exhibition of research work.The main research work in this paper promote the correlation coefficient of projection based on the results of the correlation coefficient of Mosaic image,and research the intensity-geometric correlation coefficient and correlation coefficient of diffusion.Through the different descriptions of the image form and the image features,establishing various forms of correlation coefficient.The establishment of intensity-geometric correlation coefficient of the image is based on the stochastic model and the statistic image as a tool,and is proposed and researched a series of new forms.Using the ideas of the data reduction and diffusion mapping in this paper,the method of Laplacian feature mapping and the diffusion mapping are used in image data,through researching related problems of the image heat kernel and the diffusion kernel,proposed and established the diffusion morphologic operator and applied its basic framework for image similarity comparison,proved that projection morphological operator and its correlation coefficient of Pyt'ev morphological analysis theory is a special case of diffusion morphological operator and diffusion coefficient coefficient in mosaic image mode;at the same time,by introducing the average Hausdorff distance,the proposed realization algorithm of concrete method and parameter selection methods were improved and through related experimental data shows the effectiveness of the method.