Texture Primitive Unit Extraction using Different Wavelet Transforms for Texture Classification

Texture can be characterized in different ways. Local texture facets are considered to be one of the useful approaches for texture analysis. Local texture facets comprise data about the texture behavior. The present approach extracts the texture primitive units (TPU) and texture primitive spectrum (TPS) for classification of the textures. The present paper derives a feature extraction algorithm based on TPS using wavelet decomposed images. The TPU describe the local-texture information for a given pixel and its 3x3 neighborhood. The occurrence spreading of TPUs is called as TPS. The present paper proposes a feature extraction algorithm based on TPS using wavelet decomposed images. The local-texture information for a given pixel and its neighborhood is characterized by the TPU's and TPUs calculated in different ways. The global textural aspect of an image is revealed by its TPS. The classification accuracy rates are compared with original and various wavelet sub-band images.