Classification of Textures Using Pixel Range Calculation Method

A good texture classification algorithm can easily detect and classify different types of objects. Several texture classification techniques are there to match the similar objects and distinguish different types of objects or surfaces. In this article an improvised version of texture classification technique called Pixel Range Calculation (PRC) technique has been presented. The PRC method provides better classification accuracy and takes less time for computation as compared to other discussed methods. In this article the proposed method has been compared with two state of art methods called Multi Fractal Spectrum (MFS) and Gliding Box Method (GBM). The textural dataset of Brodatz, CUReT (Columbia-Utrecht Reflectance and Texture Database) and Describable Textures Dataset (DTD) have been taken for experiment. From experimental result it has been observed that, the PRC method outperforms other two methods by its matching accuracy of 98.54%, 97.44% and 97.19% for Brodatz, CUReT and DTD dataset respectively. From confusion matrix it is observed that the PRC method is most accurate one for classification.