Two-dimensional Orthogonal Masks For Texture Classification

This paper describes an approach to texture classification in which each local neighborhood is transformed reversibly into an optimal minimal set of uncorrelated features. These features are found to represent local average, as well as first and second-order derivative information of the texture. Comparison with other techniques of similar complexity, using identical classification procedures, shows a definite superiority of the proposed approach. Moreover, this approach leads directly to a real-time parallel implementation.