A method for detecting scene cuts and similar pictures in a video sequence, including receiving pictures in a video sequence, extracting a set of features from two temporally consecutive pictures, computing a sum of square errors for the set of features with respect to the features of the previous picture, determining whether the error exceeds a predefined threshold and in response to the error exceeding the predefined threshold, detecting a scene change, and determining whether the error is less than a certain threshold thus detecting a similar picture has been found.