Comparative study of noise-tolerant texture classification

The effectiveness of four types of texture features are compared in performing noise-tolerant texture classification. The four types of texture features investigated are: edge separation texture features; standard cooccurrence matrix features; cooccurrence matrix features of directionally-smoothed texture samples; and the Fourier power and phase spectrum features of Liu and Jernigan (1990). All methods, except standard cooccurrence matrix features, performed well.<
>