The most significant features in visual scenes, is identified without prior training, by measuring the difficulty in finding similarities between neighbourhoods in the scene. Pixels in an area that is similar to much of the rest of the scene score low measures of visual attention. On the other hand a region that possesses many dissimilarities with other parts of the image will attract a high measure of visual attention. A trial and error process is used to find dissimilarities between parts of the image and does not require prior knowledge of the nature of the anomalies that may be present. The use of processing dependencies between pixels avoided while yet providing a straightforward parallel implementation for each pixel. Such techniques are of wide application in searching for anomalous patterns in health screening, quality control processes and in analysis of visual ergonomics for assessing the visibility of signs and advertisements. A measure of significant features can be provided to an image processor in order to provide variable rate image compression.