Analysis of texture: Practice

Computer-based texture analysis (TA) is used to characterize the spatial distribution of signal intensity variations within local regions (2D or 3D) in an image, and can enable us to detect signal signatures and biology-related information that are otherwise imperceptible to the human eye. Although visual texture1 is not a completely well-defined concept, image texture can be captured quantitatively using a broad range of statistical and mathematical measures. Early works by Tamura et al. (1978) attempted to develop computational definitions of basic textural features corresponding to human visual perception, and defined six basic textural properties: coarseness, contrast, directionality, line-likeness, regularity, and roughness. Since the seminal investigations in the 1970s and 1980s on texture perception in human vision by Bela Julesz at Bell Labs, there has been three main research tracks related to visual texture: (i) psychophysical investigations of visual texture perception, (ii) modeling, synthesis and mapping of texture in computer graphics for adding realism to computer-generated scenes, and (iii) texture analysis in computer vision and pattern recognition, including ‘texture classification’, ‘texture segmentation’, and ‘shape from texture’, with applications in remote sensing, industrial inspection, materials science, content-based image & video retrieval, and biological & medical imaging. In medicine, magnetic resonance imaging is potentially one of the most exciting imaging technologies where TA can be applied2. This is because of the good soft tissue contrast offered by MRI along with extensive experimental control of the spatial resolution, signal to noise ratio, and contrast mechanism during an imaging experiment or patient examination. For these reasons, there have been reported an increasing number of MRI studies where TA is used to discriminate between patient groups, or detect architectural or pathophysiological differences between normal and abnormal tissue types or tissue conditions (cf. suggested reading).