An optimal window for image analysis is determined by means of information theory. The size of the window is then used as a good estimator of image quality. The method is sensitive to the size of the structures present in the image (contrary to the contrast method, widely used to select the best images of a series) and can provide a quality map, indicating the zones more or less degraded in the image. In this letter I describe the optimal window method (OWM) and compare it with the contrast method, showing a first application to a series of images taken at Themis IPM.