Influence of magnification on feature-specific image analysis measurements

Abstract

Feature-specific measurements of test circles of known size were conducted at different magnifications using a stored image of the detected circles (256 × 256 picture points within the guard frame). Measurement of the diameter and perimeter of calibrated test circles with different objectives has demonstrated that the diameter and perimeter can be measured within an accuracy of ±2% of the true value if the particular circle is magnified to such an extent that it is 15 mm or greater in diameter on the screen. Measurement of the areas of these circles has demonstrated that the area can be measured within an accuracy of ±4% of the true value if the circle is magnified to such an extent that it is 20 mm or greater in diameter on the screen. Repeat measurements have shown that precisions greater than these (±2% for the diameter and perimeter and ±4% for the area) should not be expected. Although a 2 % error in the diameter corresponds to a 4 percent error in the area, the minimum critical sizes for the circle on the screen were not the same.

The effect of erosion and dilation procedures varied substantially with magnification. As expected, erosions and dilations have a greater influence on the measured area than the measured diameter or perimeter. The magnitude of the shift in area, diameter, and perimeter was greater for erosions than for dilations.

Measurements of the area and lengths of a large number of similar-sized squares revealed the same trends regarding the influence of the size of the square on the screen and the measurement accuracy. While the mean values for the area and length were essentially identical using × 16, × 32 and × 50 objectives, accurate definition of the dispersion of the area and length data required use of the × 50 objective for the length and either the × 32 or × 50 objective for the area, with preference for the latter.