The Application Of Pattern Classification In Relic Image Processing

Image Classification is classifying images into different classes with various image features, that is to judge which class the sample belongs in by its features. It is one aspect of pattern recognition. In this paper, we focus on the application of the pattern recognition in archaic china relic images, we describe the architecture of pattern recognition, theories, ideas and methods, and make a deep study on it in relic Image Processing.1.Clustering algorithms have been studied and three algorithms have been implemented: The hard K means, the fuzzy K means and semi-fuzzy K means algorithms. After the discussion the fuzzy K means is considered to be better for relic images classification than the other two.2.A combination of GLCM(Gray Level Co-occurrence Matrix) and Fractal features has been used: With the result of experiment we conclude that the combination feature is better than single feature.3.GMRF(Gauss Markov random field ) model based window size estimation approach for texture analysis has been improved and implemented: We use the GMRF model to describe texture, the least square error approach is used to estimate the parameters. The result shows us that it is good at describing texture.4.The invariant feature of relic images has been extracted: This signature provide a compact description of all image aspects, including color-, texture and shape, also the signature is invariant to 2D rigid transformation such as rotatiom scaling and translation.5.We have developed a real relic image classification system (particularly for archaic china relic image), based on the above.