A method of creating a 3-dimensional point cloud representation of a buoyant diffusion flame from video footage from multiple angles has been created. This point cloud contains probabilities that the flame existed over the period the method was applied. This thesis uses these methods to determine flame height, flame pulsation frequency and flame volume from video footage of two furniture fires and one pool fire. The flame height of the furniture and pool fires is compared to flame height predictions based on the heat release rate and estimated flame base diameter. The determined ‘continuous’ and ‘intermittent’ flame regions for the pool fire were always below the levels suggested by one of the predicting methods. The flame pulsation frequency was compared to a widely used correlation based on the base diameter. This equation under-predicted the pulsation frequency grossly for the furniture fires and slightly for the pool fire. It was found that by increasing the period over which the pulsation frequency was determined pulsation frequency was lowered. The flame volume was determined by summing the number of points in the point cloud representation and multiplying by the volume represented by each point. The flame volumes were compared to the experimental heat release rate from furniture calorimetry measurements to give an approximate correlation between the flame volume and the heat release rate. The values given by the method are in the same order of magnitude as given by other researchers, however insufficient video footage has been analysed to confirm the accuracy of the method. It was hoped that the methods presented could be applied to fires outside of the laboratory, however attempts to use the methods presented here on video footage of a house fire were unsuccessful due to the poor quality of the footage taken