Image Analysis (IA)
Technique Description
Computer based image analysis consists of sample selection and preparation,
image processing, measurement, and data analysis and output. Since the systems
in the Center use a standard TV signal (NTSC) from a color video camera
as an input signal, images can be provided from diverse sources. Optical
microscopes are effective for small samples while larger samples are handled
by a macroviewer. In addition, digital SEM and other images may be imported
and analyzed.Since the primary mode of image analysis is grey level thresholding or color detection, simple, high-contrast images are most appropriate for image analysis. Selecting the threshold to detect only grey levels or RGB values within specific ranges enables selective detection of features. Detection is the process which extracts a simplified, binary image from the original image.
Regardless of the input source, most images require image enhancement to generate optimal analysis results. Image processing refers to the manipulation of the detected image, for example, separating joined features or filling in voids to improve the accuracy of measurements.
Measurements are then carried out on processed binary images. The image analysis systems support two modes of measurements - field and feature. Field measurements give one total value for each parameter in an image. Some field parameters include area fraction, area, perimeter, mean chord, and corner count. Feature measurements give one individual value per parameter for each isolated object in the image. Some feature measurements include count, area, X-Y coordinates, perimeter, equivalent diameter, shape factors and grey variance.
Data may be presented in a series of histograms or scatter plots. Distributions of any measured parameter against the count or any other chosen measured parameter can be created. Calculation of histogram statistics is also possible.
Quantification of morphological aspects of imagesQuantitative determination of feature size, including particle
size distribution in powdersDigital image enhancement through processing procedures
to emphasize features of interestColor image documentation by video and laser printerPresentation of analysis results available in either tabular
form or histograms, including distribution statisticsSamples can be anything from microscopic to macroscopic regardless of their conductive nature. Images of smaller samples can be obtained through various optical or scanning electron microscopes, while larger samples may be imaged by macrolenses.
Appropriate images of the sample are acquired. The contrast and brightness of the image is adjusted until the highest possible contrast is achieved. Features of interest are detected by specifying specific grey level thresholds or RGB values. The detected image is enhanced to improve the accuracy of the measurements, then measurements are taken on the processed image.
Measures two-dimensional geometric quantities; three dimensional
parameters must be inferredDoes not provide direct chemical information on microstructural
features and cannot generally discriminate between microstructural features
of different compositionsErrors can arise in analysis measurements from images with
poor contrast. The contrast in the image must contain the information to
be measured Back to SCIC home page.