We present a new computer vision method for measuring the physical characteristics of maize silages based on the extraction of image textural features, namely five Haralick features, the moments of order 1 and 2 on the grey levels of the image, and a fractal parameter. To show the effectiveness of these features for characterising the physical properties of the maize silage, a trial was performed on three maize cultivars (Zea mays L, a mid-early hybrid and an early hybrid) to study the effects of the ensiling process. The features could discern effects of the ensiling process on the physical properties of the maize silage. These effects were neither significant by comparing particle size distributions, nor with chemical compositions, whereas a biological response (degradation in the rumen) to the ensiling process exists. Thus image textural features seemed to give new and interesting measurements of the physical properties of the silage, explaining the biological response better than other methods. Furthermore, the ensiling effect was not the same with all the maize types and varied with the proportion of large particles in the silages. Linear relations between features before and after the ensiling process made it possible to predict a measure on a silage knowing its value on the parent forage.