The aim of the study was, to use the coke images to estimate the fractal dimensions. The experiment was carried out for qualitatively diverse samples of metallurgical coke and semicokes. The basic action is to transform the colorful microscopic images of optical texture to a binary form. The obtained images are digitally processed and fractal dimensions are calculated according to box counting method. Fractal dimensions are in the range 1 < D-F <= 2. One of the main subjects was to adjust the scale of studied images of the optical texture, fora reliable assessment of fractal analysis. The small image samples (segments) of individual optical texture types were used to determine the fractal dimension. Furthermore, a method of averaging fractal dimensions of individual textures was proposed to obtain a dimensionless value, called the multifractal, D-MF. Comparisons using regression analyses between the obtained multifractal dimensions, the coke reflectance and Nippon Steel Corporation parameters were applied to check the ability of DMF to characterize the quality of coke. It was found that the correlation coefficients between DMF and selected parameters of coke quality (R-max, R-bi, CRI, CSR) are sufficiently high (0.80 <= r(2) <= 0.91). The higher the multifractal dimension of a coke is, the worse its quality will be and higher its reactivity. A very good linear correlation means that the multifractal dimension of texture can be a proper parameter for coke/semicoke quality characterization. The greatest advantage of DMF is its resistance to changes in the coking process as well as the type and quality of the input material, something that was evidenced from the examination of samples with a wide range of quality. (C) 2017 Elsevier Ltd. All rights reserved.