Texture classification by multiresolution character and fuzzy decision

A new texture classification approach by multiresolution character and fuzzy decision is proposed. The multiresolution character of the texture is extracted by a completed tree-structured wavelet transform. In the feature classification process, the advantages of the statistical and fuzzy method are combined by introducing the feature similarity measure and classifying the samples by fuzzy decision of the similarity values. The introduced similarity measure reflects the uncertainty of the feature values made by the difference between samples or made by random noise, and the proposed approach can give a satisfactory classification results. The texture classification experiment gives its effectiveness