Similarity distance measures are used to study the similarity between patterns. We propose the use of similarity measures between images to estimate the quality of combustion in a furnace designed for carbonization processes in the production of activated carbon. Broadly speaking, the production of activated carbon requires two thermal processes: carbonization and activation. One of the most sensitive variables in both processes is the level of oxygen. For carbonization, the process involves thermal decomposition of vegetal material in the absence of air. For activation, the gasification of the material at high temperature is required, and one of the oxidizing agents used is oxygen. Given the complexity of measuring the oxygen level because of the functional characteristics of the furnaces, we propose a strategy for estimating the quality of combustion, which is directly related to the oxygen level, based on similarity measures between reference photographs and the flame states. This strategy corresponds to the instrumentalization of methods used by operators in manual control of the furnaces. Our algorithm is tested with reference photos taken at the production plant, and the experimental results prove the efficiency of the proposed technique.