Multiple resolution imagery and texture analysis

摘要

A method that classifies textures at different resolutions is developed in this research. The statistical features used for classification are derived from the Fourier power spectrum and cooccurrence matrices. The relationships between feature properties and images resolution are studied. A decision process based upon “confidence levels” is used to select the correct resolution within the proper class. The experimental results indicate the effectiveness of the classification process on several texture classes with the added dimension of resolution variation.