Texture description using statistical feature extraction
Marouane Ben Haj Ayech, Hamid Amiri · 2016
Texture description becomes nowadays very important for the understanding of the visual content of the images. Several approaches are proposed in the last decades and are generally categorized into two large families: statistical and Structural. In this paper, we are interested in statistical methods which are often presented in different ways. Particularly, we propose a unified statistical approach in which we present the following techniques: Intensity histogram (IH), Gray-Level Co-occurrence Matrix (GLCM), Gray-Level Difference (GLD), Gray-Level Run Length Matrix (GLRLM) and Local Binary Pattern (LBP). Furthermore, we evaluate the ability of these methods for the task of texture description using a dedicated challenging benchmark DTD. The experiments show that LBP outperforms the other methods.