In this paper we present a heuristic perceptually-based termination criterion for stochastic radiosity algorithms. The proposed criterion makes it possible to terminate the iterative radiosity simulation as soon as any further changes of the radiosity solution are predicted not to be noticed by the human observer. We use tone mapping operators and perceptually uniform CIE L*u*v* colour space to arrive at the difference of displayed values. In order to decide when to stop the iterations of the radiosity method, we evaluate the least squares fit of a power function to the error distribution data.