Recognizing the central quality of a metallographic sample represents, for a metallographic laboratory expert, a challenging task involving analyzing the size and number of defects at the evaluated segment. In order to help human experts, a software tool was developed for machine evaluation of these metallographic samples. The aim of the paper is to present the properties of the developed software and to demonstrate its practical application in industry. The developed software tool is based on the machine recognition of small but important non-homogeneity objects located at the central part of the metallographic sample and on a statistical evaluation of the properties of these extracted central objects. Because the central objects are very similar to noise, machine recognition is not straightforward. However, this paper contains promising results of comparing machine and expert assessments of an industrial image database of digitized metallographic samples. The image database is provided by the research department of ArcelorMittal Ostrava, a.s., in the Czech Republic.