We have studied the usefulness of the AWS saliency maps for proto-objects and context extraction from images. With this aim, we have performed a simple (and non-parametric) watershed segmentation of the obtained saliency maps with different images with very different scenarios and under very different lighting conditions.