Pronouncedly inhomogeneous thermograms occur universally in infrared imaging applications. The heterogeneity as a separate characteristic of a 2D pattern merits detailed consideration and quantitative estimation. The present contribution is among the first works dedicated to solving this problem. In addition to the review and summarizing of previous results, a new computational method is developed for quantitative estimation of the thermal pattern inhomogeneity. Artificial and really measured thermograms are analyzed with the designed original software. An adequacy of the new method is confirmed.