In this contribution a comparison of two approaches for classification of metallography images from the steel plant of Mittal Steel Ostrava plc (Ostrava, Czech Republic) is presented. The aim of the classification is to monitor the process quality in the steel plant. The first classifier represents images by feature vectors extracted using the wavelet transformation, while the feature computation in the second approach is based on the eigen-space analysis. Experiments made for real metallography data indicate feasibility of both methods for automatic image classification in hard industry environment.