An inverse radiation analysis is presented for determining the three-dimensional temperature field in an inhomogeneous, absorbing, emitting and anisotropically scattering media of known radiative properties from the knowledge of the exit radiative energy received by charge-coupled device (CCD) cameras at boundary surfaces. The forward Monte Carlo method was employed to describe the radiative energy propagation. The inverse problem was formulated as an ill-posed matrix equation and solved by least square QR decomposition (LSQR) method. The measured data were simulated by adding random errors to the exact solution of the direct problem. The effects of measurement errors, combinations of CCD cameras, concentration distributions of particles, and coefficient fluctuating errors on the accuracy of the inverse problem were investigated. The results show that the three-dimensional temperature field call be estimated accurately, even for the noisy data. (c) 2007 Published by Elsevier Ltd.