Abstract: Computational vision-based fire and flame detection has drawn significant attention in the past decade with camera surveillance systems becoming ubiquitous. Signal and image processing methods are developed for the detection of fire, flames and smoke in open and large spaces with a range of up to 30m to the camera in visible-range (IR) video. This paper proposes a new approach to computational vision-based fire and flame detection by using a compound algorithm and a decision fusion framework with Naïve Bayes classifier as classification tool. The compound algorithm consists of several sub-algorithms, the fusion network is to fuse the results obtained by each of these sub-algorithms and Naïve Bayes classifier is useful for the final classification. This approach is to improve the accuracy of fire and flame detection in videos and to reduce the false alarm rate to a great extent.