The leaf area index (LAI) of forest canopies can be estimated indirectly by measuring their light transmission, either with a specific sensor or by hemispherical photographs. The angle at which the transmission is measured enters the calculations at two different points: (1) as angle of incidence determining the travel distance of a light ray through the whole canopy and (2) relatively to the zenith for the statistical distribution of the angle at which the single foliage elements are seen. On a flat ground, any direction is sufficiently described by its zenith angle. On a slope, however, the angle of incidence is not identical to the zenith angle. Each direction in the canopy has thus to be classified according to both angles. We propose here an iterative method to introduce this slope effect into the calculation of LAI (and mean leaf angle) from hemispherical pictures. This method was tested both on artificial pictures simulating ideal canopies and on photographs taken in forests of different LAI, growing on slopes up to 37°. This method was found to give accurate results, while neglecting the effect of the slope would lead to underestimations of LAI, especially in dense canopies. On a flat ground, the proposed method also improves the accuracy of the calculations because it better accounts for the non-linear relationship between light transmission and LAI.