Flashes of light from impacts in Jupiter can be detected by running differential photometry over a video observation. A reference frame is taken at the beginning of the video sequence and later frames are compared with it so that differential photometry of the whole planet is computed. The reference can be updated over the video or computed from a number of coregistered frames improving the signal to noise ratio of the reference image. Then each frame is compared with a reference image and substracted so that a faint impact appears as a bright signal over a background of noise. The different software tools that you will find in this page use different implementations of this basic idea. This idea was proposed simultaneously by Emil Kraaikamp and Ricardo Hueso and is also the basis for the extraction of light-curves of the impact that allow to characterize the energy released.
This is demonstraded in the image shown on the right where the spot covered by the impact flash departs clearly from the rest of the planet. This example corresponds to the first impact flash on June 3, 2010. In order to compare frames they need to be coregistered. A simple algorithm based on finding the Center of Brightness (CB) position can be easily implemented. This also requires an estimation of the planet size in the video from horizontal and vertical scan centered on the CB pixel.
The software assumes an impact departs strongly from the rest of the planet and looks for the frames with the pixels with the uppermost differential photometric value. The software orders the frames in terms of the strength of the most intense pixel and compares them sequentially identifying those consecutive frames where pixels with the strongest signal are found. Note that due to the seeing effects the planet limb contributes strongly to the signal of the differential photometry.
All software tools in these pages use different implementations of this idea.
An additional concept, developed by Marc Delcroix, is to coregister all the images in a video frame and consider the brightest value of a pixel over a video sequence and the average value for each position. The difference between these two images offers the capability to detect the fainter photometric signal of less intense bolides or bolides observed with modest equipment.
The algorithm consists on the following steps: