Quantitative Image Analysis

Computer-based image analysis is a contrast-based microstructural analysis tool that is finding widespread use because of significant improvements in computer speed and memory capabilities. The modern quantitative image analysis systems driven by personal computers can handle complex images quickly and accurately, with resolutions comparable to those of the human eye. The most economical image systems convert an analog camera signal to a digital signal through a hardware processor. Although digital cameras are available, they are significantly more expensive.

Image analysis consists of six basic steps (Ref 2):

1. Image acquisition for microstructural analysis is accomplished with a microscope and a video camera.

2. Image enhancement is required to correct for nonuniform illumination and to sharpen the digitized image. Image sharpening is required because the conversion from an analog signal to a digitized signal does not produce sharp edges. Delineation, the most commonly used image enhancement function, converts digitized signals to a square-wave function, thus producing sharper images.

3. Feature detection (also called thresholding) assigns specific contrast values to the image (0 to 255 gray levels). The most accurate and precise method for properly thresholding an image is to use a gray-level histogram. Figure 1 shows a porous steatite ceramic sample and the corresponding gray-level histogram. The darker phase (porosity) occurs at a lower value and is detected between gray levels of 0 to 129; the lighter phase (background matrix) is detected between gray levels of 130 to 255.

4. Editing is used to separate, delineate, or classify features based on their morphology or size. Examples of editing functions include dilation, erosion, grain-boundary reconstruction, filing, elimination, and chord sizing. These functions are used to add or remove pixels, rebuild missing or incomplete boundaries, fill in undetected features, or eliminate objects greater than a set pixel size. Another useful editing feature is the use of Boolean operators, such as AND, OR, XOR, NOR, NXOR, and NAND, to combine or separate features.

5. Quantitative measurement and analysis is performed for three categories of data. Object measurements analyze individual features such as area, length, width, aspect ratio, and sphericity (see Table 1). Field measurements analyze individual objects and report values over the entire field. Examples include area fraction, area percent, and density (see Table 1). Total area measurements analyze the cumulative values reported over multiple fields. Figure 2 shows the analysis of an etched steatite sample. Note that the phase percentages as well as the percent porosity have been measured.

6. Report generation presents the data in a clear and understandable fashion.

Fig. 1 Gray-level threshold for a porous steatite ceramic sample
Fig. 2 Phase analysis of an etched steatite ceramic sample
0 0

Post a comment