Texture Recognition Based on Invariant Analysis and Support Vector Machine

摘要

Texture recognition plays an important role in machine vision and signal processing. Over the last three decades, many texture analysis and recognition methods have been proposed, but most of them are sensitive to geometric distortion. This paper presents a new approach to scaling and rotation invariant texture classification using invariant moments based on radon transform. The radon transform is utilized to project the texture image onto projection space. Then the geometric invariant moments are calculated by the projection data. The support vector machine-based classifier is employed to implement the texture classification. Experimental results show the high classification accuracy of this approach.