This paper presents an overview of methods of error elimination during image processing and
analysis, and more specially ofthe problems ofsetting the threshold errors, shading and noise.
Of all the steps involved in measuring objects in an image, setting the threshold is the one
which should be considered most careftilly. There are described following errors: threshold setting
error, the offset threshold error, "halo" error and elimination methods.
The process of setting the detection level should be standardised, whichever technique is
adopted, and some form of check on a standard specimen should be incorporated into the image
analysis routine if at all practicable.
Shading is the best dealt with by using a shading correction step in the image processing, but
operator can help by ensuring that illumination system of microscope for example is correctly
centred, that specimen is flat, that the lamps on macroviewer all have the same wattage and are at
the same distance from the specimen. Shading can be asymmetrical, and it can be present in the
specimen itself as well as being introduced in the optical system.
Noise is a function of system design, with no room for operator of the system to affect it,
except by increasing the gain of the video amplification. Noise is minimised by operating the
system at the optimum conditions of incident lighting.