The Research Of Granular Computing Theory And Its Application In Image Enhancing

From different levels of granules, Granular Computing (GrC) studies problems which are imprise, partial truth, fuzzy and uncountable. As a new research field, "GrC"has two special tasks. One task is to obtain their common characteristics of different field without consideration of their difference in lower levels, in order to get their abstract, higher levels' and global comprehension. The other is to crystallize the connotative structures of given field in order to conclude the universal principles which do not depend on one field. The unique charactistic is that the complicated problems are solved by systematic and structural comprehension and method. The comprehension of complicated problem is usually multi-perspective, and the comprehension from each perspective is multi-level. So, the process of "GrC" just is the process of solving complicated problem. Its result is a multi-perspective and multi-level granular structure which is the systematic and approximate solution of complicated problem.As the subset of "GrC", rough set theory study the problems which are imintegral or uncertain. The idea of classification is the theoretical basis of data processing with rough set theory. The abstract meaning of classification is the basis of learning and decision. The main advantage of rough set theory is that it does not need any transcendental and additional knowledge.This paper has studied some theory about "GrC" based on rough set theory. A reduction algorithm was proposed in this paper based on set approximation in "GrC". The attribute reduction was processed according to the approximation so that the decision table can be reduced. In the paper the fuzzy concepts of fuzzy theory were represented by binary numeral. Then, the fuzzy logic operation proceeds according to binary numeral granule, which obtains better result of reasoning. There are many advantages in the process of a larage number of data and the lost information based on "GrC", so "GrC' was applied in data mining system in this paper. The combination of the technology of data mining and "GrC" can reduce the influence of the supposed parameter in mathematical model of traditional data mining. The decision rules was easy to understand based on "GrC", which is also proposed in the thesis.The main task of image enhancing operation is to highlight the information interested and weakens or eliminates the unnessary information in order to enhancing the usful information. The purpose of image enhancing is: on one hand, the vision effect of image was improved with a serial technics to improve the articulation of image. On the other hand, the image was translated into a form that is better suit for man or machine to explain and analyze. Image enhancing is to highlight the information instead of image fidelity so that man or machine can analyze the interested information with processing the image and restrain the unusful information in order to improve the value of use, that is, image enhancing just enhances discrimination of some information.Integrality and uncertainty may occer in many different levels in the process of image because of the complication and the strong correlation of image information. The application of "GrC" in image process and comprehension can get better result than hard computing method. The image was considered as a knowledge system in this paper, the image system was divided into different sub-block according to undistinct relationship. Then, each sub-block was enhanced. At last, conclude the whole result of each sub-block to get the result of image enhancing based on "GrC".The image enhancing program was programmed and was tested by MATLAB, and the proposed "GrC" method was applied in colorful image in this paper.The experiment approves that the colorful image with noises would become distinct after the process of "GrC".In a word, "GrC" is a new method of information process, which can be used in many different fields. It is of great comprehension and strong vitality. It is a method that can solve problem more easily.