A general method for the construction of the optimum classical description
of a physical process on the atomic scale is presented. The method is devel
oped for physical systems whose quantum versions are obtained by canonical
quantization of their classical counterparts and it is in principle applica
ble even in the case of 'low' quantum numbers. The criterion determining th
e optimum description was essentially based on the comparison of the quantu
m and classical probability distributions in terms of the canonical coordin
ates. The main application of the method is in classical calculations of va
rious quantities measured in atomic collision processes.