An analysis of a thermal image (thermogram) of a human body surface shows the availability of correlation between the body state and the thermogram heterogeneity (“motley”). A new method for the quantitative analysis of the thermogram heterogeneity based on the thermogram representation by a quadtree of reduced images is proposed. In this method, we partition any image fragment into four parts when this fragment is heterogeneous. The image heterogeneity is evaluated as a value proportional to the number of leaves of the constructed quadtree. For amplification of the heterogeneity degree dependence upon an observed “motley” of real thermograms, a well known contrast expansion is applied.