Classification of Textures by Avoiding Complex Patterns

Abstract

The present research advocates a new method for texture classification based on primitive patterns. For this, an optimal set of eight simple patterns are selected based on their discrimination. The percentage of frequency of occurrences is computed for these patterns on entire texture image. Textures can be classified by using frequency of occurrences, for complex patterns also. The present research proposes a scheme to completely avoid the computation of, frequency of occurrence for complex patterns. Further, an innovative scheme for determining the percentage of occurrences of upper and lower limits of complex patterns is focused in the present study. The experimental results proved the avoidance of computation of complex patterns. Finally, the experimental results on Brodatz textures indicate the clear classification of textures into classes by selecting simple patterns, but the classes are overlapped when complex patterns are applied.