Currently digital video are widely used in various fields of science and technology and in human daily activities. Intensively growing and already existing huge amount of digital video data need to be managed, so the shot boundary detection is the first and important step for content-based video retrieval and indexing. Each algorithm aimed for this approach should accurately detect boundaries between camera shots and do a segmentation of a video. In this paper a new method of abrupt transitions detection is proposed, based on Weibull distribution of each frame's gradient magnitude. Experimental results successfully show that it can effectively detect hard cuts and has certain advantages against widely used other methods, which are using image pixels point-by-point comparison method.