In this study, digital image processing is employed as a discriminant method for wide areas and the results of the examination of digital images of various road surface conditions procured in the laboratory experiment and in the field are presented. The following conclusions were obtained from the results of the laboratory and the on-site experiments on discriminant methods of winter road conditions based on image processing: (1) The results of the laboratory and on-site experiments showed that three statistical parameters (contrast, skewness and kurtosis) were effective in discriminating between road surface texture of single road surface conditions. However, an autocorrelation function and a power spectrum generated by Fourier transform do not clearly show a difference in road conditions; (2) The discrimination between complex road surfaces is sometimes not clear because clusters of parameters of each road surface are adjacent. However, by applying fuzzy theory, the percentage of true prediction reaches about 80 per cent; and (3) Although a wavelet analysis was performed to examine a method for discriminating between local parts of road surfaces, it could simply make the location of a different road surface condition clear. It is necessary to discriminate also between road surface conditions in the future, by giving a more detailed multiple resolution analysis of the original signal.