Digital image analysis is a method of extraction of useful information from digital images by using various digital image processing techniques. A digital image is a two-dimensional matrix that consists of great number of smallest addressable elements – pixels. Each pixel has its own digital characteristics, such as address and color representation. Digital image characteristics have two main measures: spatial resolution that is focused on capturing details and tonal resolution that is used to measure color, dynamic range and bit-depth. Higher tonal, spatial, and color values mean larger sizes of digital image file. (Peterson, 2005, pp. 1-2)
The appropriate method is needed to utilize color as a visual cue in digital image analysis. Color models (also called color solids or color spaces) are used to represent a color signal. A correctly chosen color model preserves essential information about the digital image and grants insight to the needed visual operation. In order to choose the best color representation, the image analyst should know the process of color signal generation and what kind of information he/she needs to receive. (Plataniotis & Venetsanopoulos, 2000, p. 3)
All color models are based on either additive or subtractive approach. The most commonly used additive color system is RGB (Red-Green-Blue). It is supported by majority of software programs and devices. The primary colors used in subtractive color models are cyan, magenta and yellow. These pigments absorb more light if they are added to a white page. Common subtractive color models are the CIE-L*A*B and CMYK. (Rhyne, 2012, pp. 3-6,11)
Digital image classification uses the quantitative spectral information that is contained in an image. Such information is represented by digital data in spectral bands. As result of digital image classification, each individual pixel carries information about different spectral bands. Analysis can be performed on both multispectral and hyperspectral imagery. The objective of image classification is to assign each pixel in the image to a specific theme or class. A class could be water, corn, coniferous forest, rice and many others. Initial results have raster format, but may be further processed to polygons. Results of such analysis can be used in GIS systems. There are three main principles of digital image analysis: spectral differentiation, radiometric differentiation and spatial differentiation. According to spectral differentiation, objects that have different condition and nature give different colors in an image. Another principle, radiometric differentiation is based on the detection of differences in brightness. Spatial differentiation uses different resolution and it may be used to measure the size of the objects.
There are two main approaches to digital image classification: unsupervised and supervised. Amount and nature of human interaction needed are different for each approach. In supervised classification the image analyst needs to choose a classification scheme and identify different surface types of interest (called training areas). Accurate results of such approach depend mainly on ability of the analyst. There is much less input required from the analyst if using unsupervised classification approach. The classification algorithm analyses the image itself, combining pixels into clusters. The analyst must determine if the algorithm has picked the classes correctly. (Schuckman & Dutton, 2014)
Today, digital image analysis is used in various fields of science and engineering. For example, a group of British scientists successfully applied digital image processing techniques to quantify submerged deposited fine sediments in rivers. Such analysis may be used in various stream and river research, management and restoration projects because it provides ecologically friendly and non-destructive approach. (Turley et al., 2016, p. 9) Digital image analysis can also enable geologists to study structural features, such as lineaments and lithologies more effectively. Researcher from Kuwait Al-Shumaimri used such approach to research geological and geomorphological features of Jordan. (Al-Shumaimri, 2012, p. 41)
References
Al-Shumaimri, M. (2012). Application of Digital Image Processing Techniques to Geological and Geomorphological Features of Southwest Jordan. Journal of Geography and Geology, 4(1), 41-48.
Peterson, K. A. (2005). Introduction to Basic Measures of a Digital Image for Pictorial Collections. 8. https://www.loc.gov/rr/print/tp/IntroDgtlImage.pdf
Plataniotis, K., & Venetsanopoulos, A. N. (2000). Color image processing and applications: Springer Science & Business Media.
Rhyne, M. T. (2012). Applying color theory to digital media & visualization (pp. 91).
Schuckman, K., & Dutton, J. A. (2014). Digital Image Classification. Retrieved 01.25.2017, from https://www.e-education.psu.edu/geog480/node/496
Turley, M., Bilotta, G., Arbociute, G., Chadd, R., Extence, C., & Brazier, R. (2016). Quantifying Submerged Deposited Fine Sediments in Rivers and Streams Using Digital Image Analysis. River Research and Applications, 11.