Image Analysis

Dimensional quantification of sample microcharacteristics is a frequent customer need. Whether it be thickness of oxide layers on a corroded steel or the size distribution of the starting materials for a powder metallurgy process, dimensional analysis plays an important role in understanding how to deal with sample failure or process design. MATCO has advanced capabilities to provide this kind of information to our customers.

Many software suppliers claim to make this sort of analysis a "snap." Twenty years of experience working with it says it doesn't work that way! The reality is that each sample presents its own special problems for quantification and that attempting to use a strictly software-based approach will eat up hundreds of fruitless hours. It is far better to use a software-enhanced human-eye-based approach, grounded in deep understanding of the materials being examined.

An example of the difficulty encountered in a strictly software based approach is the problem of determining size distributions in oil-in-water micro-emulsions. After solving the problem of obtaining images of this relatively high-vapor-pressure system, one will find that the images of each droplet contain a range of gray scales that overlaps the range in the image area between the droplets. Thus no computer-based system of segregating the droplets on the basis of gray level will be able to avoid including background in the selected area. It is simply impossible. It may be possible to segregate the droplets based on shape (i.e., they should be spherical), but again, normally the droplets will be severely aggregated, and the time spent figuring out how to teach the computer to disaggregate them efficiently and perfectly could be better spent doing the size analysis by a rapid computer-assisted manual method.

The complexity of this problem is matched in a large fraction of the routine problems encountered in SEM work. It has usually proven more efficient to do size distributions by some at least partly subjective manual approach than to try to do it totally objectively. A major problem with laser-based systems, for example, particularly in the submicron range, is not knowing to what degree the sample is aggregated. If the human eye is involved, it can sort out that problem.

Please bring your real-world image analysis problems to us at MATCO and we will provide you with experienced real-world answers.