In automatic and image processing based wildfire smoke detection various spatial and temporal procedures are used to detect certain wildfire features. The final step is usually the classification procedure that finally decides if some detected phenomena represents wildfire smoke or not. In classification procedure various classifiers can be used, but as the system has to work in real time and usually on limited hardware, the efforts of researchers working on those systems are usually aimed towards the implementation of classifiers that are simple for implementation and fast in execution. In this paper the emphasis is on the comparison of two simple and fast classifiers — the cogent confabulation based classifier and the Naive Bayes classifier. Even though they can appear similar, on the theoretical and philosophical level they are quite different. Besides the theoretical comparison of both classifiers they are also compared on an example of lens flare detection in images taken from wildfire smoke surveillance cameras video streams.