To have knowledge about lightning detection is an important aspect in development of protection systems to provide security to living beings and assure the correct performance of electronic and electrical devices in electrical storms. Actually, there are technologies to lightning detection based on magnetic and electromagnetic fields measurements which do not always provide information regarding of event occurrence. By using cameras in the process to acquire lightning images and video will serve as a redundant system in lightning detection to the existing ones. In order to determine whether or not the object detected visually corresponds to a lightning, a shape descriptor may be used. Because lightning images have an elongated, thin and polygonal shape, this paper proposes a descriptor of lightning without branches through the application of tortuosity metric. The method to find the appropriate descriptor begins with the measurement of tortuosity indexes to a set of segmented images of lightning without branches with the purpose of selecting the appropriate range for each tortuosity index. Afterwards, the measurements of tortuosity indexes are used in two image sets with the objective of evaluating the descriptor performance, one corresponding to lightning images and the second set with images associated with other natural phenomena such as rainbow, light reflections in clouds and on water surface, crepuscular rays and wind waves. The obtained results are promising, giving an accuracy of 86.11% with the combination of tortuosity indexes &ñ964;2 and &ñ964;4.