Grey level co-occurrence matrix and its application to seismic data

T exture analysis is the extraction of textural features from images (Tuceryan and Jain, 1998). The meaning of texture varies, depending on the area of science in which it is used. In general, texture refers to the physical character of an object or the appearance of an image. In image analysis, texture is defined as a function of the spatial variation in intensities of pixels (Tuceryan and Jain, 1998). Seismic texture refers to the magnitude and variability of neighbouring amplitudes at sample locations and is physically related to the distribution of scattering objects (geological texture) within a small volume at the corresponding subsurface location (Gao, 2008). Four principal methods have been developed for the analysis of seismic texture (Figure 1). These are texture classification, segmentation, synthesis, and shape. The aim of texture classification is to categorize features in an image by recognizing known texture classes. This approach is easy to compute and is the most used method of texture analysis. Texture segmentation partitions an image into different regions that have homogeneous properties. The segmentation can be either based on regions or based on boundaries between regions. In texture synthesis, small sample images are used as the basis for the construction or reconstruction of larger images. This methodology is widely applied in the reconstruction of digital images and in postproduction of films. Texture shape is the least used method of texture analysis. It uses texture information to construct 3D surface geometries. According to Tuceryan and Jain (1998), texture classification can be divided into four computational categories: statistical, geometrical, model-based, and signal processing methods of computation. Numerous applications are available that employ each of these methods. Statistical texture analysis, such as the grey level cooccurrence matrix (GLCM), grey level differences, or local binary pattern extraction, try to define the arrangement of different regions in an image through statistics. Statistical methods do not attempt to understand the hierarchical